Commit | Line | Data |
---|---|---|
a0d0e21e LW |
1 | /* regexec.c |
2 | */ | |
3 | ||
4 | /* | |
4ac71550 TC |
5 | * One Ring to rule them all, One Ring to find them |
6 | & | |
7 | * [p.v of _The Lord of the Rings_, opening poem] | |
8 | * [p.50 of _The Lord of the Rings_, I/iii: "The Shadow of the Past"] | |
9 | * [p.254 of _The Lord of the Rings_, II/ii: "The Council of Elrond"] | |
a0d0e21e LW |
10 | */ |
11 | ||
61296642 DM |
12 | /* This file contains functions for executing a regular expression. See |
13 | * also regcomp.c which funnily enough, contains functions for compiling | |
166f8a29 | 14 | * a regular expression. |
e4a054ea DM |
15 | * |
16 | * This file is also copied at build time to ext/re/re_exec.c, where | |
17 | * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT. | |
18 | * This causes the main functions to be compiled under new names and with | |
19 | * debugging support added, which makes "use re 'debug'" work. | |
166f8a29 DM |
20 | */ |
21 | ||
a687059c LW |
22 | /* NOTE: this is derived from Henry Spencer's regexp code, and should not |
23 | * confused with the original package (see point 3 below). Thanks, Henry! | |
24 | */ | |
25 | ||
26 | /* Additional note: this code is very heavily munged from Henry's version | |
27 | * in places. In some spots I've traded clarity for efficiency, so don't | |
28 | * blame Henry for some of the lack of readability. | |
29 | */ | |
30 | ||
e50aee73 AD |
31 | /* The names of the functions have been changed from regcomp and |
32 | * regexec to pregcomp and pregexec in order to avoid conflicts | |
33 | * with the POSIX routines of the same names. | |
34 | */ | |
35 | ||
b9d5759e | 36 | #ifdef PERL_EXT_RE_BUILD |
54df2634 | 37 | #include "re_top.h" |
9041c2e3 | 38 | #endif |
56953603 | 39 | |
7e0d5ad7 KW |
40 | /* At least one required character in the target string is expressible only in |
41 | * UTF-8. */ | |
42 | const char* const non_utf8_target_but_utf8_required | |
43 | = "Can't match, because target string needs to be in UTF-8\n"; | |
44 | ||
a687059c | 45 | /* |
e50aee73 | 46 | * pregcomp and pregexec -- regsub and regerror are not used in perl |
a687059c LW |
47 | * |
48 | * Copyright (c) 1986 by University of Toronto. | |
49 | * Written by Henry Spencer. Not derived from licensed software. | |
50 | * | |
51 | * Permission is granted to anyone to use this software for any | |
52 | * purpose on any computer system, and to redistribute it freely, | |
53 | * subject to the following restrictions: | |
54 | * | |
55 | * 1. The author is not responsible for the consequences of use of | |
56 | * this software, no matter how awful, even if they arise | |
57 | * from defects in it. | |
58 | * | |
59 | * 2. The origin of this software must not be misrepresented, either | |
60 | * by explicit claim or by omission. | |
61 | * | |
62 | * 3. Altered versions must be plainly marked as such, and must not | |
63 | * be misrepresented as being the original software. | |
64 | * | |
65 | **** Alterations to Henry's code are... | |
66 | **** | |
4bb101f2 | 67 | **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, |
1129b882 NC |
68 | **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 |
69 | **** by Larry Wall and others | |
a687059c | 70 | **** |
9ef589d8 LW |
71 | **** You may distribute under the terms of either the GNU General Public |
72 | **** License or the Artistic License, as specified in the README file. | |
a687059c LW |
73 | * |
74 | * Beware that some of this code is subtly aware of the way operator | |
75 | * precedence is structured in regular expressions. Serious changes in | |
76 | * regular-expression syntax might require a total rethink. | |
77 | */ | |
78 | #include "EXTERN.h" | |
864dbfa3 | 79 | #define PERL_IN_REGEXEC_C |
a687059c | 80 | #include "perl.h" |
0f5d15d6 | 81 | |
54df2634 NC |
82 | #ifdef PERL_IN_XSUB_RE |
83 | # include "re_comp.h" | |
84 | #else | |
85 | # include "regcomp.h" | |
86 | #endif | |
a687059c | 87 | |
81e983c1 | 88 | #include "inline_invlist.c" |
1b0f46bf | 89 | #include "unicode_constants.h" |
81e983c1 | 90 | |
ef07e810 | 91 | #define RF_tainted 1 /* tainted information used? e.g. locale */ |
c277df42 | 92 | #define RF_warned 2 /* warned about big count? */ |
faec1544 | 93 | |
ab3bbdeb | 94 | #define RF_utf8 8 /* Pattern contains multibyte chars? */ |
a0ed51b3 | 95 | |
f2ed9b32 | 96 | #define UTF_PATTERN ((PL_reg_flags & RF_utf8) != 0) |
ce862d02 | 97 | |
c74f6de9 KW |
98 | #define HAS_NONLATIN1_FOLD_CLOSURE(i) _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i) |
99 | ||
a687059c LW |
100 | #ifndef STATIC |
101 | #define STATIC static | |
102 | #endif | |
103 | ||
e0193e47 | 104 | /* Valid for non-utf8 strings: avoids the reginclass |
7e2509c1 KW |
105 | * call if there are no complications: i.e., if everything matchable is |
106 | * straight forward in the bitmap */ | |
af364d03 KW |
107 | #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) \ |
108 | : ANYOF_BITMAP_TEST(p,*(c))) | |
7d3e948e | 109 | |
c277df42 IZ |
110 | /* |
111 | * Forwards. | |
112 | */ | |
113 | ||
f2ed9b32 | 114 | #define CHR_SVLEN(sv) (utf8_target ? sv_len_utf8(sv) : SvCUR(sv)) |
53c4c00c | 115 | #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b) |
a0ed51b3 | 116 | |
3dab1dad YO |
117 | #define HOPc(pos,off) \ |
118 | (char *)(PL_reg_match_utf8 \ | |
52657f30 | 119 | ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \ |
3dab1dad YO |
120 | : (U8*)(pos + off)) |
121 | #define HOPBACKc(pos, off) \ | |
07be1b83 YO |
122 | (char*)(PL_reg_match_utf8\ |
123 | ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \ | |
124 | : (pos - off >= PL_bostr) \ | |
8e11feef | 125 | ? (U8*)pos - off \ |
3dab1dad | 126 | : NULL) |
efb30f32 | 127 | |
e7409c1b | 128 | #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off)) |
1aa99e6b | 129 | #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim)) |
1aa99e6b | 130 | |
7016d6eb DM |
131 | |
132 | #define NEXTCHR_EOS -10 /* nextchr has fallen off the end */ | |
133 | #define NEXTCHR_IS_EOS (nextchr < 0) | |
134 | ||
135 | #define SET_nextchr \ | |
136 | nextchr = ((locinput < PL_regeol) ? UCHARAT(locinput) : NEXTCHR_EOS) | |
137 | ||
138 | #define SET_locinput(p) \ | |
139 | locinput = (p); \ | |
140 | SET_nextchr | |
141 | ||
142 | ||
20d0b1e9 | 143 | /* these are unrolled below in the CCC_TRY_XXX defined */ |
61dad979 | 144 | #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \ |
9c4fdda1 | 145 | if (!CAT2(PL_utf8_,class)) { \ |
cf54f63a | 146 | bool ok; \ |
9c4fdda1 | 147 | ENTER; save_re_context(); \ |
cf54f63a JL |
148 | ok=CAT2(is_utf8_,class)((const U8*)str); \ |
149 | PERL_UNUSED_VAR(ok); \ | |
150 | assert(ok); assert(CAT2(PL_utf8_,class)); LEAVE; } } STMT_END | |
37e2e78e KW |
151 | /* Doesn't do an assert to verify that is correct */ |
152 | #define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \ | |
9c4fdda1 | 153 | if (!CAT2(PL_utf8_,class)) { \ |
9d63fa07 | 154 | bool throw_away PERL_UNUSED_DECL; \ |
9c4fdda1 RB |
155 | ENTER; save_re_context(); \ |
156 | throw_away = CAT2(is_utf8_,class)((const U8*)" "); \ | |
157 | LEAVE; } } STMT_END | |
37e2e78e | 158 | |
1a4fad37 AL |
159 | #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a") |
160 | #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0") | |
161 | #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ") | |
51371543 | 162 | |
37e2e78e | 163 | #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \ |
61dad979 KW |
164 | /* No asserts are done for some of these, in case called on a */ \ |
165 | /* Unicode version in which they map to nothing */ \ | |
27d4fc33 | 166 | LOAD_UTF8_CHARCLASS(X_regular_begin, HYPHEN_UTF8); \ |
61dad979 | 167 | LOAD_UTF8_CHARCLASS(X_extend, COMBINING_GRAVE_ACCENT_UTF8); \ |
20d0b1e9 | 168 | |
1dcf4a1b | 169 | #define PLACEHOLDER /* Something for the preprocessor to grab onto */ |
d1eb3177 | 170 | |
ee9a90b8 KW |
171 | /* The actual code for CCC_TRY, which uses several variables from the routine |
172 | * it's callable from. It is designed to be the bulk of a case statement. | |
173 | * FUNC is the macro or function to call on non-utf8 targets that indicate if | |
174 | * nextchr matches the class. | |
175 | * UTF8_TEST is the whole test string to use for utf8 targets | |
176 | * LOAD is what to use to test, and if not present to load in the swash for the | |
177 | * class | |
178 | * POS_OR_NEG is either empty or ! to complement the results of FUNC or | |
179 | * UTF8_TEST test. | |
180 | * The logic is: Fail if we're at the end-of-string; otherwise if the target is | |
181 | * utf8 and a variant, load the swash if necessary and test using the utf8 | |
182 | * test. Advance to the next character if test is ok, otherwise fail; If not | |
183 | * utf8 or an invariant under utf8, use the non-utf8 test, and fail if it | |
184 | * fails, or advance to the next character */ | |
185 | ||
186 | #define _CCC_TRY_CODE(POS_OR_NEG, FUNC, UTF8_TEST, CLASS, STR) \ | |
7016d6eb | 187 | if (NEXTCHR_IS_EOS) { \ |
ee9a90b8 KW |
188 | sayNO; \ |
189 | } \ | |
190 | if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \ | |
191 | LOAD_UTF8_CHARCLASS(CLASS, STR); \ | |
192 | if (POS_OR_NEG (UTF8_TEST)) { \ | |
193 | sayNO; \ | |
194 | } \ | |
ee9a90b8 | 195 | } \ |
28b98f76 DM |
196 | else if (POS_OR_NEG (FUNC(nextchr))) { \ |
197 | sayNO; \ | |
ee9a90b8 | 198 | } \ |
28b98f76 | 199 | goto increment_locinput; |
980866de | 200 | |
ee9a90b8 KW |
201 | /* Handle the non-locale cases for a character class and its complement. It |
202 | * calls _CCC_TRY_CODE with a ! to complement the test for the character class. | |
203 | * This is because that code fails when the test succeeds, so we want to have | |
204 | * the test fail so that the code succeeds. The swash is stored in a | |
205 | * predictable PL_ place */ | |
cfaf538b KW |
206 | #define _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, \ |
207 | CLASS, STR) \ | |
ee9a90b8 KW |
208 | case NAME: \ |
209 | _CCC_TRY_CODE( !, FUNC, \ | |
210 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ | |
211 | (U8*)locinput, TRUE)), \ | |
212 | CLASS, STR) \ | |
213 | case NNAME: \ | |
1dcf4a1b | 214 | _CCC_TRY_CODE( PLACEHOLDER , FUNC, \ |
ee9a90b8 KW |
215 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ |
216 | (U8*)locinput, TRUE)), \ | |
217 | CLASS, STR) \ | |
218 | ||
219 | /* Generate the case statements for both locale and non-locale character | |
220 | * classes in regmatch for classes that don't have special unicode semantics. | |
221 | * Locales don't use an immediate swash, but an intermediary special locale | |
222 | * function that is called on the pointer to the current place in the input | |
223 | * string. That function will resolve to needing the same swash. One might | |
224 | * think that because we don't know what the locale will match, we shouldn't | |
225 | * check with the swash loading function that it loaded properly; ie, that we | |
226 | * should use LOAD_UTF8_CHARCLASS_NO_CHECK for those, but what is passed to the | |
227 | * regular LOAD_UTF8_CHARCLASS is in non-locale terms, and so locale is | |
228 | * irrelevant here */ | |
229 | #define CCC_TRY(NAME, NNAME, FUNC, \ | |
230 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
cfaf538b | 231 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 KW |
232 | CLASS, STR) \ |
233 | case NAMEL: \ | |
234 | PL_reg_flags |= RF_tainted; \ | |
235 | _CCC_TRY_CODE( !, LCFUNC, LCFUNC_utf8((U8*)locinput), CLASS, STR) \ | |
236 | case NNAMEL: \ | |
237 | PL_reg_flags |= RF_tainted; \ | |
1dcf4a1b KW |
238 | _CCC_TRY_CODE( PLACEHOLDER, LCFUNC, LCFUNC_utf8((U8*)locinput), \ |
239 | CLASS, STR) \ | |
cfaf538b | 240 | case NAMEA: \ |
7016d6eb | 241 | if (NEXTCHR_IS_EOS || ! FUNCA(nextchr)) { \ |
cfaf538b KW |
242 | sayNO; \ |
243 | } \ | |
244 | /* Matched a utf8-invariant, so don't have to worry about utf8 */ \ | |
3640db6b | 245 | locinput++; \ |
cfaf538b KW |
246 | break; \ |
247 | case NNAMEA: \ | |
7016d6eb | 248 | if (NEXTCHR_IS_EOS || FUNCA(nextchr)) { \ |
cfaf538b KW |
249 | sayNO; \ |
250 | } \ | |
28b98f76 | 251 | goto increment_locinput; \ |
ee9a90b8 KW |
252 | /* Generate the non-locale cases */ \ |
253 | _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, CLASS, STR) | |
254 | ||
255 | /* This is like CCC_TRY, but has an extra set of parameters for generating case | |
256 | * statements to handle separate Unicode semantics nodes */ | |
257 | #define CCC_TRY_U(NAME, NNAME, FUNC, \ | |
258 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
259 | NAMEU, NNAMEU, FUNCU, \ | |
cfaf538b | 260 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 | 261 | CLASS, STR) \ |
cfaf538b KW |
262 | CCC_TRY(NAME, NNAME, FUNC, \ |
263 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
264 | NAMEA, NNAMEA, FUNCA, \ | |
265 | CLASS, STR) \ | |
ee9a90b8 | 266 | _CCC_TRY_NONLOCALE(NAMEU, NNAMEU, FUNCU, CLASS, STR) |
d1eb3177 | 267 | |
3dab1dad YO |
268 | /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */ |
269 | ||
5f80c4cf | 270 | /* for use after a quantifier and before an EXACT-like node -- japhy */ |
c35dcbe2 YO |
271 | /* it would be nice to rework regcomp.sym to generate this stuff. sigh |
272 | * | |
273 | * NOTE that *nothing* that affects backtracking should be in here, specifically | |
274 | * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a | |
275 | * node that is in between two EXACT like nodes when ascertaining what the required | |
276 | * "follow" character is. This should probably be moved to regex compile time | |
277 | * although it may be done at run time beause of the REF possibility - more | |
278 | * investigation required. -- demerphq | |
279 | */ | |
3e901dc0 YO |
280 | #define JUMPABLE(rn) ( \ |
281 | OP(rn) == OPEN || \ | |
282 | (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \ | |
283 | OP(rn) == EVAL || \ | |
cca55fe3 JP |
284 | OP(rn) == SUSPEND || OP(rn) == IFMATCH || \ |
285 | OP(rn) == PLUS || OP(rn) == MINMOD || \ | |
d1c771f5 | 286 | OP(rn) == KEEPS || \ |
3dab1dad | 287 | (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \ |
e2d8ce26 | 288 | ) |
ee9b8eae | 289 | #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT) |
e2d8ce26 | 290 | |
ee9b8eae YO |
291 | #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF ) |
292 | ||
293 | #if 0 | |
294 | /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so | |
295 | we don't need this definition. */ | |
296 | #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF ) | |
fab2782b | 297 | #define IS_TEXTF(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFU_TRICKYFOLD || OP(rn)==EXACTFA || OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF ) |
ee9b8eae YO |
298 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL ) |
299 | ||
300 | #else | |
301 | /* ... so we use this as its faster. */ | |
302 | #define IS_TEXT(rn) ( OP(rn)==EXACT ) | |
fab2782b | 303 | #define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFU_TRICKYFOLD || OP(rn) == EXACTFA) |
ee9b8eae YO |
304 | #define IS_TEXTF(rn) ( OP(rn)==EXACTF ) |
305 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL ) | |
306 | ||
307 | #endif | |
e2d8ce26 | 308 | |
a84d97b6 HS |
309 | /* |
310 | Search for mandatory following text node; for lookahead, the text must | |
311 | follow but for lookbehind (rn->flags != 0) we skip to the next step. | |
312 | */ | |
cca55fe3 | 313 | #define FIND_NEXT_IMPT(rn) STMT_START { \ |
3dab1dad YO |
314 | while (JUMPABLE(rn)) { \ |
315 | const OPCODE type = OP(rn); \ | |
316 | if (type == SUSPEND || PL_regkind[type] == CURLY) \ | |
e2d8ce26 | 317 | rn = NEXTOPER(NEXTOPER(rn)); \ |
3dab1dad | 318 | else if (type == PLUS) \ |
cca55fe3 | 319 | rn = NEXTOPER(rn); \ |
3dab1dad | 320 | else if (type == IFMATCH) \ |
a84d97b6 | 321 | rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \ |
e2d8ce26 | 322 | else rn += NEXT_OFF(rn); \ |
3dab1dad | 323 | } \ |
5f80c4cf | 324 | } STMT_END |
74750237 | 325 | |
c476f425 | 326 | |
acfe0abc | 327 | static void restore_pos(pTHX_ void *arg); |
51371543 | 328 | |
87c0511b | 329 | #define REGCP_PAREN_ELEMS 3 |
f067efbf | 330 | #define REGCP_OTHER_ELEMS 3 |
e0fa7e2b | 331 | #define REGCP_FRAME_ELEMS 1 |
620d5b66 NC |
332 | /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and |
333 | * are needed for the regexp context stack bookkeeping. */ | |
334 | ||
76e3520e | 335 | STATIC CHECKPOINT |
b93070ed | 336 | S_regcppush(pTHX_ const regexp *rex, I32 parenfloor) |
a0d0e21e | 337 | { |
97aff369 | 338 | dVAR; |
a3b680e6 | 339 | const int retval = PL_savestack_ix; |
a3b680e6 | 340 | const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; |
e0fa7e2b NC |
341 | const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS; |
342 | const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT; | |
87c0511b | 343 | I32 p; |
40a82448 | 344 | GET_RE_DEBUG_FLAGS_DECL; |
a0d0e21e | 345 | |
b93070ed DM |
346 | PERL_ARGS_ASSERT_REGCPPUSH; |
347 | ||
e49a9654 | 348 | if (paren_elems_to_push < 0) |
5637ef5b NC |
349 | Perl_croak(aTHX_ "panic: paren_elems_to_push, %i < 0", |
350 | paren_elems_to_push); | |
e49a9654 | 351 | |
e0fa7e2b NC |
352 | if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems) |
353 | Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf | |
5df417d0 JH |
354 | " out of range (%lu-%ld)", |
355 | total_elems, (unsigned long)PL_regsize, (long)parenfloor); | |
e0fa7e2b | 356 | |
620d5b66 | 357 | SSGROW(total_elems + REGCP_FRAME_ELEMS); |
7f69552c | 358 | |
495f47a5 DM |
359 | DEBUG_BUFFERS_r( |
360 | if ((int)PL_regsize > (int)parenfloor) | |
361 | PerlIO_printf(Perl_debug_log, | |
362 | "rex=0x%"UVxf" offs=0x%"UVxf": saving capture indices:\n", | |
363 | PTR2UV(rex), | |
364 | PTR2UV(rex->offs) | |
365 | ); | |
366 | ); | |
87c0511b | 367 | for (p = parenfloor+1; p <= (I32)PL_regsize; p++) { |
b1ce53c5 | 368 | /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ |
b93070ed DM |
369 | SSPUSHINT(rex->offs[p].end); |
370 | SSPUSHINT(rex->offs[p].start); | |
1ca2007e | 371 | SSPUSHINT(rex->offs[p].start_tmp); |
e7707071 | 372 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
495f47a5 DM |
373 | " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"\n", |
374 | (UV)p, | |
375 | (IV)rex->offs[p].start, | |
376 | (IV)rex->offs[p].start_tmp, | |
377 | (IV)rex->offs[p].end | |
40a82448 | 378 | )); |
a0d0e21e | 379 | } |
b1ce53c5 | 380 | /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ |
3280af22 | 381 | SSPUSHINT(PL_regsize); |
b93070ed DM |
382 | SSPUSHINT(rex->lastparen); |
383 | SSPUSHINT(rex->lastcloseparen); | |
e0fa7e2b | 384 | SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */ |
41123dfd | 385 | |
a0d0e21e LW |
386 | return retval; |
387 | } | |
388 | ||
c277df42 | 389 | /* These are needed since we do not localize EVAL nodes: */ |
ab3bbdeb YO |
390 | #define REGCP_SET(cp) \ |
391 | DEBUG_STATE_r( \ | |
ab3bbdeb | 392 | PerlIO_printf(Perl_debug_log, \ |
e4f74956 | 393 | " Setting an EVAL scope, savestack=%"IVdf"\n", \ |
ab3bbdeb YO |
394 | (IV)PL_savestack_ix)); \ |
395 | cp = PL_savestack_ix | |
c3464db5 | 396 | |
ab3bbdeb | 397 | #define REGCP_UNWIND(cp) \ |
e4f74956 | 398 | DEBUG_STATE_r( \ |
ab3bbdeb | 399 | if (cp != PL_savestack_ix) \ |
e4f74956 YO |
400 | PerlIO_printf(Perl_debug_log, \ |
401 | " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ | |
ab3bbdeb YO |
402 | (IV)(cp), (IV)PL_savestack_ix)); \ |
403 | regcpblow(cp) | |
c277df42 | 404 | |
a8d1f4b4 DM |
405 | #define UNWIND_PAREN(lp, lcp) \ |
406 | for (n = rex->lastparen; n > lp; n--) \ | |
407 | rex->offs[n].end = -1; \ | |
408 | rex->lastparen = n; \ | |
409 | rex->lastcloseparen = lcp; | |
410 | ||
411 | ||
f067efbf | 412 | STATIC void |
b93070ed | 413 | S_regcppop(pTHX_ regexp *rex) |
a0d0e21e | 414 | { |
97aff369 | 415 | dVAR; |
e0fa7e2b | 416 | UV i; |
87c0511b | 417 | U32 paren; |
a3621e74 YO |
418 | GET_RE_DEBUG_FLAGS_DECL; |
419 | ||
7918f24d NC |
420 | PERL_ARGS_ASSERT_REGCPPOP; |
421 | ||
b1ce53c5 | 422 | /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ |
c6bf6a65 | 423 | i = SSPOPUV; |
e0fa7e2b NC |
424 | assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ |
425 | i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */ | |
b93070ed DM |
426 | rex->lastcloseparen = SSPOPINT; |
427 | rex->lastparen = SSPOPINT; | |
3280af22 | 428 | PL_regsize = SSPOPINT; |
b1ce53c5 | 429 | |
620d5b66 | 430 | i -= REGCP_OTHER_ELEMS; |
b1ce53c5 | 431 | /* Now restore the parentheses context. */ |
495f47a5 DM |
432 | DEBUG_BUFFERS_r( |
433 | if (i || rex->lastparen + 1 <= rex->nparens) | |
434 | PerlIO_printf(Perl_debug_log, | |
435 | "rex=0x%"UVxf" offs=0x%"UVxf": restoring capture indices to:\n", | |
436 | PTR2UV(rex), | |
437 | PTR2UV(rex->offs) | |
438 | ); | |
439 | ); | |
87c0511b | 440 | paren = PL_regsize; |
620d5b66 | 441 | for ( ; i > 0; i -= REGCP_PAREN_ELEMS) { |
1df70142 | 442 | I32 tmps; |
1ca2007e | 443 | rex->offs[paren].start_tmp = SSPOPINT; |
b93070ed | 444 | rex->offs[paren].start = SSPOPINT; |
cf93c79d | 445 | tmps = SSPOPINT; |
b93070ed DM |
446 | if (paren <= rex->lastparen) |
447 | rex->offs[paren].end = tmps; | |
495f47a5 DM |
448 | DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log, |
449 | " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"%s\n", | |
450 | (UV)paren, | |
451 | (IV)rex->offs[paren].start, | |
452 | (IV)rex->offs[paren].start_tmp, | |
453 | (IV)rex->offs[paren].end, | |
454 | (paren > rex->lastparen ? "(skipped)" : "")); | |
c277df42 | 455 | ); |
87c0511b | 456 | paren--; |
a0d0e21e | 457 | } |
daf18116 | 458 | #if 1 |
dafc8851 JH |
459 | /* It would seem that the similar code in regtry() |
460 | * already takes care of this, and in fact it is in | |
461 | * a better location to since this code can #if 0-ed out | |
462 | * but the code in regtry() is needed or otherwise tests | |
463 | * requiring null fields (pat.t#187 and split.t#{13,14} | |
daf18116 JH |
464 | * (as of patchlevel 7877) will fail. Then again, |
465 | * this code seems to be necessary or otherwise | |
225593e1 DM |
466 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ |
467 | * --jhi updated by dapm */ | |
b93070ed | 468 | for (i = rex->lastparen + 1; i <= rex->nparens; i++) { |
097eb12c | 469 | if (i > PL_regsize) |
b93070ed DM |
470 | rex->offs[i].start = -1; |
471 | rex->offs[i].end = -1; | |
495f47a5 DM |
472 | DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log, |
473 | " \\%"UVuf": %s ..-1 undeffing\n", | |
474 | (UV)i, | |
475 | (i > PL_regsize) ? "-1" : " " | |
476 | )); | |
a0d0e21e | 477 | } |
dafc8851 | 478 | #endif |
a0d0e21e LW |
479 | } |
480 | ||
74088413 DM |
481 | /* restore the parens and associated vars at savestack position ix, |
482 | * but without popping the stack */ | |
483 | ||
484 | STATIC void | |
485 | S_regcp_restore(pTHX_ regexp *rex, I32 ix) | |
486 | { | |
487 | I32 tmpix = PL_savestack_ix; | |
488 | PL_savestack_ix = ix; | |
489 | regcppop(rex); | |
490 | PL_savestack_ix = tmpix; | |
491 | } | |
492 | ||
02db2b7b | 493 | #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ |
a0d0e21e | 494 | |
a687059c | 495 | /* |
e50aee73 | 496 | * pregexec and friends |
a687059c LW |
497 | */ |
498 | ||
76234dfb | 499 | #ifndef PERL_IN_XSUB_RE |
a687059c | 500 | /* |
c277df42 | 501 | - pregexec - match a regexp against a string |
a687059c | 502 | */ |
c277df42 | 503 | I32 |
49d7dfbc | 504 | Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend, |
c3464db5 | 505 | char *strbeg, I32 minend, SV *screamer, U32 nosave) |
8fd1a950 DM |
506 | /* stringarg: the point in the string at which to begin matching */ |
507 | /* strend: pointer to null at end of string */ | |
508 | /* strbeg: real beginning of string */ | |
509 | /* minend: end of match must be >= minend bytes after stringarg. */ | |
510 | /* screamer: SV being matched: only used for utf8 flag, pos() etc; string | |
511 | * itself is accessed via the pointers above */ | |
512 | /* nosave: For optimizations. */ | |
c277df42 | 513 | { |
7918f24d NC |
514 | PERL_ARGS_ASSERT_PREGEXEC; |
515 | ||
c277df42 | 516 | return |
9041c2e3 | 517 | regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, |
c277df42 IZ |
518 | nosave ? 0 : REXEC_COPY_STR); |
519 | } | |
76234dfb | 520 | #endif |
22e551b9 | 521 | |
9041c2e3 | 522 | /* |
cad2e5aa JH |
523 | * Need to implement the following flags for reg_anch: |
524 | * | |
525 | * USE_INTUIT_NOML - Useful to call re_intuit_start() first | |
526 | * USE_INTUIT_ML | |
527 | * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer | |
528 | * INTUIT_AUTORITATIVE_ML | |
529 | * INTUIT_ONCE_NOML - Intuit can match in one location only. | |
530 | * INTUIT_ONCE_ML | |
531 | * | |
532 | * Another flag for this function: SECOND_TIME (so that float substrs | |
533 | * with giant delta may be not rechecked). | |
534 | */ | |
535 | ||
536 | /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */ | |
537 | ||
3f7c398e | 538 | /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend. |
cad2e5aa JH |
539 | Otherwise, only SvCUR(sv) is used to get strbeg. */ |
540 | ||
541 | /* XXXX We assume that strpos is strbeg unless sv. */ | |
542 | ||
6eb5f6b9 JH |
543 | /* XXXX Some places assume that there is a fixed substring. |
544 | An update may be needed if optimizer marks as "INTUITable" | |
545 | RExen without fixed substrings. Similarly, it is assumed that | |
546 | lengths of all the strings are no more than minlen, thus they | |
547 | cannot come from lookahead. | |
40d049e4 YO |
548 | (Or minlen should take into account lookahead.) |
549 | NOTE: Some of this comment is not correct. minlen does now take account | |
550 | of lookahead/behind. Further research is required. -- demerphq | |
551 | ||
552 | */ | |
6eb5f6b9 | 553 | |
2c2d71f5 JH |
554 | /* A failure to find a constant substring means that there is no need to make |
555 | an expensive call to REx engine, thus we celebrate a failure. Similarly, | |
556 | finding a substring too deep into the string means that less calls to | |
30944b6d IZ |
557 | regtry() should be needed. |
558 | ||
559 | REx compiler's optimizer found 4 possible hints: | |
560 | a) Anchored substring; | |
561 | b) Fixed substring; | |
562 | c) Whether we are anchored (beginning-of-line or \G); | |
486ec47a | 563 | d) First node (of those at offset 0) which may distinguish positions; |
6eb5f6b9 | 564 | We use a)b)d) and multiline-part of c), and try to find a position in the |
30944b6d IZ |
565 | string which does not contradict any of them. |
566 | */ | |
2c2d71f5 | 567 | |
6eb5f6b9 JH |
568 | /* Most of decisions we do here should have been done at compile time. |
569 | The nodes of the REx which we used for the search should have been | |
570 | deleted from the finite automaton. */ | |
571 | ||
cad2e5aa | 572 | char * |
288b8c02 | 573 | Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos, |
9f61653a | 574 | char *strend, const U32 flags, re_scream_pos_data *data) |
cad2e5aa | 575 | { |
97aff369 | 576 | dVAR; |
288b8c02 | 577 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
eb578fdb | 578 | I32 start_shift = 0; |
cad2e5aa | 579 | /* Should be nonnegative! */ |
eb578fdb KW |
580 | I32 end_shift = 0; |
581 | char *s; | |
582 | SV *check; | |
a1933d95 | 583 | char *strbeg; |
cad2e5aa | 584 | char *t; |
f2ed9b32 | 585 | const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */ |
cad2e5aa | 586 | I32 ml_anch; |
eb578fdb | 587 | char *other_last = NULL; /* other substr checked before this */ |
bd61b366 | 588 | char *check_at = NULL; /* check substr found at this pos */ |
d8080198 | 589 | char *checked_upto = NULL; /* how far into the string we have already checked using find_byclass*/ |
bbe252da | 590 | const I32 multiline = prog->extflags & RXf_PMf_MULTILINE; |
f8fc2ecf | 591 | RXi_GET_DECL(prog,progi); |
30944b6d | 592 | #ifdef DEBUGGING |
890ce7af | 593 | const char * const i_strpos = strpos; |
30944b6d | 594 | #endif |
a3621e74 YO |
595 | GET_RE_DEBUG_FLAGS_DECL; |
596 | ||
7918f24d | 597 | PERL_ARGS_ASSERT_RE_INTUIT_START; |
c33e64f0 FC |
598 | PERL_UNUSED_ARG(flags); |
599 | PERL_UNUSED_ARG(data); | |
7918f24d | 600 | |
f2ed9b32 | 601 | RX_MATCH_UTF8_set(rx,utf8_target); |
cad2e5aa | 602 | |
3c8556c3 | 603 | if (RX_UTF8(rx)) { |
b8d68ded JH |
604 | PL_reg_flags |= RF_utf8; |
605 | } | |
ab3bbdeb | 606 | DEBUG_EXECUTE_r( |
f2ed9b32 | 607 | debug_start_match(rx, utf8_target, strpos, strend, |
1de06328 YO |
608 | sv ? "Guessing start of match in sv for" |
609 | : "Guessing start of match in string for"); | |
2a782b5b | 610 | ); |
cad2e5aa | 611 | |
c344f387 JH |
612 | /* CHR_DIST() would be more correct here but it makes things slow. */ |
613 | if (prog->minlen > strend - strpos) { | |
a3621e74 | 614 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 | 615 | "String too short... [re_intuit_start]\n")); |
cad2e5aa | 616 | goto fail; |
2c2d71f5 | 617 | } |
1de06328 | 618 | |
7016d6eb DM |
619 | /* XXX we need to pass strbeg as a separate arg: the following is |
620 | * guesswork and can be wrong... */ | |
621 | if (sv && SvPOK(sv)) { | |
622 | char * p = SvPVX(sv); | |
623 | STRLEN cur = SvCUR(sv); | |
624 | if (p <= strpos && strpos < p + cur) { | |
625 | strbeg = p; | |
626 | assert(p <= strend && strend <= p + cur); | |
627 | } | |
628 | else | |
629 | strbeg = strend - cur; | |
630 | } | |
631 | else | |
632 | strbeg = strpos; | |
633 | ||
1aa99e6b | 634 | PL_regeol = strend; |
f2ed9b32 | 635 | if (utf8_target) { |
33b8afdf JH |
636 | if (!prog->check_utf8 && prog->check_substr) |
637 | to_utf8_substr(prog); | |
638 | check = prog->check_utf8; | |
639 | } else { | |
7e0d5ad7 KW |
640 | if (!prog->check_substr && prog->check_utf8) { |
641 | if (! to_byte_substr(prog)) { | |
642 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
643 | non_utf8_target_but_utf8_required)); | |
644 | goto fail; | |
645 | } | |
646 | } | |
33b8afdf JH |
647 | check = prog->check_substr; |
648 | } | |
bbe252da YO |
649 | if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */ |
650 | ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE) | |
651 | || ( (prog->extflags & RXf_ANCH_BOL) | |
7fba1cd6 | 652 | && !multiline ) ); /* Check after \n? */ |
cad2e5aa | 653 | |
7e25d62c | 654 | if (!ml_anch) { |
bbe252da YO |
655 | if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */ |
656 | && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */ | |
3f7c398e | 657 | /* SvCUR is not set on references: SvRV and SvPVX_const overlap */ |
7e25d62c JH |
658 | && sv && !SvROK(sv) |
659 | && (strpos != strbeg)) { | |
a3621e74 | 660 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); |
7e25d62c JH |
661 | goto fail; |
662 | } | |
663 | if (prog->check_offset_min == prog->check_offset_max && | |
bbe252da | 664 | !(prog->extflags & RXf_CANY_SEEN)) { |
2c2d71f5 | 665 | /* Substring at constant offset from beg-of-str... */ |
cad2e5aa JH |
666 | I32 slen; |
667 | ||
1aa99e6b | 668 | s = HOP3c(strpos, prog->check_offset_min, strend); |
1de06328 | 669 | |
653099ff GS |
670 | if (SvTAIL(check)) { |
671 | slen = SvCUR(check); /* >= 1 */ | |
cad2e5aa | 672 | |
9041c2e3 | 673 | if ( strend - s > slen || strend - s < slen - 1 |
2c2d71f5 | 674 | || (strend - s == slen && strend[-1] != '\n')) { |
a3621e74 | 675 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n")); |
2c2d71f5 | 676 | goto fail_finish; |
cad2e5aa JH |
677 | } |
678 | /* Now should match s[0..slen-2] */ | |
679 | slen--; | |
3f7c398e | 680 | if (slen && (*SvPVX_const(check) != *s |
cad2e5aa | 681 | || (slen > 1 |
3f7c398e | 682 | && memNE(SvPVX_const(check), s, slen)))) { |
2c2d71f5 | 683 | report_neq: |
a3621e74 | 684 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); |
2c2d71f5 JH |
685 | goto fail_finish; |
686 | } | |
cad2e5aa | 687 | } |
3f7c398e | 688 | else if (*SvPVX_const(check) != *s |
653099ff | 689 | || ((slen = SvCUR(check)) > 1 |
3f7c398e | 690 | && memNE(SvPVX_const(check), s, slen))) |
2c2d71f5 | 691 | goto report_neq; |
c315bfe8 | 692 | check_at = s; |
2c2d71f5 | 693 | goto success_at_start; |
7e25d62c | 694 | } |
cad2e5aa | 695 | } |
2c2d71f5 | 696 | /* Match is anchored, but substr is not anchored wrt beg-of-str. */ |
cad2e5aa | 697 | s = strpos; |
2c2d71f5 | 698 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
1de06328 YO |
699 | end_shift = prog->check_end_shift; |
700 | ||
2c2d71f5 | 701 | if (!ml_anch) { |
a3b680e6 | 702 | const I32 end = prog->check_offset_max + CHR_SVLEN(check) |
653099ff | 703 | - (SvTAIL(check) != 0); |
a3b680e6 | 704 | const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; |
2c2d71f5 JH |
705 | |
706 | if (end_shift < eshift) | |
707 | end_shift = eshift; | |
708 | } | |
cad2e5aa | 709 | } |
2c2d71f5 | 710 | else { /* Can match at random position */ |
cad2e5aa JH |
711 | ml_anch = 0; |
712 | s = strpos; | |
1de06328 YO |
713 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
714 | end_shift = prog->check_end_shift; | |
715 | ||
716 | /* end shift should be non negative here */ | |
cad2e5aa JH |
717 | } |
718 | ||
bcdf7404 | 719 | #ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */ |
0033605d | 720 | if (end_shift < 0) |
1de06328 | 721 | Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ", |
220fc49f | 722 | (IV)end_shift, RX_PRECOMP(prog)); |
2c2d71f5 JH |
723 | #endif |
724 | ||
2c2d71f5 JH |
725 | restart: |
726 | /* Find a possible match in the region s..strend by looking for | |
727 | the "check" substring in the region corrected by start/end_shift. */ | |
1de06328 YO |
728 | |
729 | { | |
730 | I32 srch_start_shift = start_shift; | |
731 | I32 srch_end_shift = end_shift; | |
c33e64f0 FC |
732 | U8* start_point; |
733 | U8* end_point; | |
1de06328 YO |
734 | if (srch_start_shift < 0 && strbeg - s > srch_start_shift) { |
735 | srch_end_shift -= ((strbeg - s) - srch_start_shift); | |
736 | srch_start_shift = strbeg - s; | |
737 | } | |
6bda09f9 | 738 | DEBUG_OPTIMISE_MORE_r({ |
1de06328 YO |
739 | PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n", |
740 | (IV)prog->check_offset_min, | |
741 | (IV)srch_start_shift, | |
742 | (IV)srch_end_shift, | |
743 | (IV)prog->check_end_shift); | |
744 | }); | |
745 | ||
bbe252da | 746 | if (prog->extflags & RXf_CANY_SEEN) { |
1de06328 YO |
747 | start_point= (U8*)(s + srch_start_shift); |
748 | end_point= (U8*)(strend - srch_end_shift); | |
749 | } else { | |
750 | start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend); | |
751 | end_point= HOP3(strend, -srch_end_shift, strbeg); | |
752 | } | |
6bda09f9 | 753 | DEBUG_OPTIMISE_MORE_r({ |
56570a2c | 754 | PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n", |
1de06328 | 755 | (int)(end_point - start_point), |
fc8cd66c | 756 | (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point), |
1de06328 YO |
757 | start_point); |
758 | }); | |
759 | ||
760 | s = fbm_instr( start_point, end_point, | |
7fba1cd6 | 761 | check, multiline ? FBMrf_MULTILINE : 0); |
1de06328 | 762 | } |
cad2e5aa JH |
763 | /* Update the count-of-usability, remove useless subpatterns, |
764 | unshift s. */ | |
2c2d71f5 | 765 | |
ab3bbdeb | 766 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 767 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
768 | SvPVX_const(check), RE_SV_DUMPLEN(check), 30); |
769 | PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s", | |
2c2d71f5 | 770 | (s ? "Found" : "Did not find"), |
f2ed9b32 | 771 | (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) |
ab3bbdeb YO |
772 | ? "anchored" : "floating"), |
773 | quoted, | |
774 | RE_SV_TAIL(check), | |
775 | (s ? " at offset " : "...\n") ); | |
776 | }); | |
2c2d71f5 JH |
777 | |
778 | if (!s) | |
779 | goto fail_finish; | |
2c2d71f5 | 780 | /* Finish the diagnostic message */ |
a3621e74 | 781 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); |
2c2d71f5 | 782 | |
1de06328 YO |
783 | /* XXX dmq: first branch is for positive lookbehind... |
784 | Our check string is offset from the beginning of the pattern. | |
785 | So we need to do any stclass tests offset forward from that | |
786 | point. I think. :-( | |
787 | */ | |
788 | ||
789 | ||
790 | ||
791 | check_at=s; | |
792 | ||
793 | ||
2c2d71f5 JH |
794 | /* Got a candidate. Check MBOL anchoring, and the *other* substr. |
795 | Start with the other substr. | |
796 | XXXX no SCREAM optimization yet - and a very coarse implementation | |
a0288114 | 797 | XXXX /ttx+/ results in anchored="ttx", floating="x". floating will |
2c2d71f5 JH |
798 | *always* match. Probably should be marked during compile... |
799 | Probably it is right to do no SCREAM here... | |
800 | */ | |
801 | ||
f2ed9b32 | 802 | if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8) |
1de06328 YO |
803 | : (prog->float_substr && prog->anchored_substr)) |
804 | { | |
30944b6d | 805 | /* Take into account the "other" substring. */ |
2c2d71f5 JH |
806 | /* XXXX May be hopelessly wrong for UTF... */ |
807 | if (!other_last) | |
6eb5f6b9 | 808 | other_last = strpos; |
f2ed9b32 | 809 | if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) { |
30944b6d IZ |
810 | do_other_anchored: |
811 | { | |
890ce7af AL |
812 | char * const last = HOP3c(s, -start_shift, strbeg); |
813 | char *last1, *last2; | |
be8e71aa | 814 | char * const saved_s = s; |
33b8afdf | 815 | SV* must; |
2c2d71f5 | 816 | |
2c2d71f5 JH |
817 | t = s - prog->check_offset_max; |
818 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ | |
f2ed9b32 | 819 | && (!utf8_target |
0ce71af7 | 820 | || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos)) |
2c2d71f5 | 821 | && t > strpos))) |
6f207bd3 | 822 | NOOP; |
2c2d71f5 JH |
823 | else |
824 | t = strpos; | |
1aa99e6b | 825 | t = HOP3c(t, prog->anchored_offset, strend); |
6eb5f6b9 JH |
826 | if (t < other_last) /* These positions already checked */ |
827 | t = other_last; | |
1aa99e6b | 828 | last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); |
2c2d71f5 JH |
829 | if (last < last1) |
830 | last1 = last; | |
1de06328 YO |
831 | /* XXXX It is not documented what units *_offsets are in. |
832 | We assume bytes, but this is clearly wrong. | |
833 | Meaning this code needs to be carefully reviewed for errors. | |
834 | dmq. | |
835 | */ | |
836 | ||
2c2d71f5 | 837 | /* On end-of-str: see comment below. */ |
f2ed9b32 | 838 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; |
33b8afdf JH |
839 | if (must == &PL_sv_undef) { |
840 | s = (char*)NULL; | |
1de06328 | 841 | DEBUG_r(must = prog->anchored_utf8); /* for debug */ |
33b8afdf JH |
842 | } |
843 | else | |
844 | s = fbm_instr( | |
845 | (unsigned char*)t, | |
846 | HOP3(HOP3(last1, prog->anchored_offset, strend) | |
847 | + SvCUR(must), -(SvTAIL(must)!=0), strbeg), | |
848 | must, | |
7fba1cd6 | 849 | multiline ? FBMrf_MULTILINE : 0 |
33b8afdf | 850 | ); |
ab3bbdeb | 851 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 852 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
853 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
854 | PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s", | |
2c2d71f5 | 855 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
856 | quoted, RE_SV_TAIL(must)); |
857 | }); | |
858 | ||
859 | ||
2c2d71f5 JH |
860 | if (!s) { |
861 | if (last1 >= last2) { | |
a3621e74 | 862 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 JH |
863 | ", giving up...\n")); |
864 | goto fail_finish; | |
865 | } | |
a3621e74 | 866 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 | 867 | ", trying floating at offset %ld...\n", |
be8e71aa | 868 | (long)(HOP3c(saved_s, 1, strend) - i_strpos))); |
1aa99e6b IH |
869 | other_last = HOP3c(last1, prog->anchored_offset+1, strend); |
870 | s = HOP3c(last, 1, strend); | |
2c2d71f5 JH |
871 | goto restart; |
872 | } | |
873 | else { | |
a3621e74 | 874 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
30944b6d | 875 | (long)(s - i_strpos))); |
1aa99e6b IH |
876 | t = HOP3c(s, -prog->anchored_offset, strbeg); |
877 | other_last = HOP3c(s, 1, strend); | |
be8e71aa | 878 | s = saved_s; |
2c2d71f5 JH |
879 | if (t == strpos) |
880 | goto try_at_start; | |
2c2d71f5 JH |
881 | goto try_at_offset; |
882 | } | |
30944b6d | 883 | } |
2c2d71f5 JH |
884 | } |
885 | else { /* Take into account the floating substring. */ | |
33b8afdf | 886 | char *last, *last1; |
be8e71aa | 887 | char * const saved_s = s; |
33b8afdf JH |
888 | SV* must; |
889 | ||
890 | t = HOP3c(s, -start_shift, strbeg); | |
891 | last1 = last = | |
892 | HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg); | |
893 | if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset) | |
894 | last = HOP3c(t, prog->float_max_offset, strend); | |
895 | s = HOP3c(t, prog->float_min_offset, strend); | |
896 | if (s < other_last) | |
897 | s = other_last; | |
2c2d71f5 | 898 | /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ |
f2ed9b32 | 899 | must = utf8_target ? prog->float_utf8 : prog->float_substr; |
33b8afdf JH |
900 | /* fbm_instr() takes into account exact value of end-of-str |
901 | if the check is SvTAIL(ed). Since false positives are OK, | |
902 | and end-of-str is not later than strend we are OK. */ | |
903 | if (must == &PL_sv_undef) { | |
904 | s = (char*)NULL; | |
1de06328 | 905 | DEBUG_r(must = prog->float_utf8); /* for debug message */ |
33b8afdf JH |
906 | } |
907 | else | |
2c2d71f5 | 908 | s = fbm_instr((unsigned char*)s, |
33b8afdf JH |
909 | (unsigned char*)last + SvCUR(must) |
910 | - (SvTAIL(must)!=0), | |
7fba1cd6 | 911 | must, multiline ? FBMrf_MULTILINE : 0); |
ab3bbdeb | 912 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 913 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
914 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
915 | PerlIO_printf(Perl_debug_log, "%s floating substr %s%s", | |
33b8afdf | 916 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
917 | quoted, RE_SV_TAIL(must)); |
918 | }); | |
33b8afdf JH |
919 | if (!s) { |
920 | if (last1 == last) { | |
a3621e74 | 921 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf JH |
922 | ", giving up...\n")); |
923 | goto fail_finish; | |
2c2d71f5 | 924 | } |
a3621e74 | 925 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf | 926 | ", trying anchored starting at offset %ld...\n", |
be8e71aa | 927 | (long)(saved_s + 1 - i_strpos))); |
33b8afdf JH |
928 | other_last = last; |
929 | s = HOP3c(t, 1, strend); | |
930 | goto restart; | |
931 | } | |
932 | else { | |
a3621e74 | 933 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
33b8afdf JH |
934 | (long)(s - i_strpos))); |
935 | other_last = s; /* Fix this later. --Hugo */ | |
be8e71aa | 936 | s = saved_s; |
33b8afdf JH |
937 | if (t == strpos) |
938 | goto try_at_start; | |
939 | goto try_at_offset; | |
940 | } | |
2c2d71f5 | 941 | } |
cad2e5aa | 942 | } |
2c2d71f5 | 943 | |
1de06328 | 944 | |
9ef43ace | 945 | t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos); |
1de06328 | 946 | |
6bda09f9 | 947 | DEBUG_OPTIMISE_MORE_r( |
1de06328 YO |
948 | PerlIO_printf(Perl_debug_log, |
949 | "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n", | |
950 | (IV)prog->check_offset_min, | |
951 | (IV)prog->check_offset_max, | |
952 | (IV)(s-strpos), | |
953 | (IV)(t-strpos), | |
954 | (IV)(t-s), | |
955 | (IV)(strend-strpos) | |
956 | ) | |
957 | ); | |
958 | ||
2c2d71f5 | 959 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ |
f2ed9b32 | 960 | && (!utf8_target |
9ef43ace | 961 | || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos))) |
1de06328 YO |
962 | && t > strpos))) |
963 | { | |
2c2d71f5 JH |
964 | /* Fixed substring is found far enough so that the match |
965 | cannot start at strpos. */ | |
966 | try_at_offset: | |
cad2e5aa | 967 | if (ml_anch && t[-1] != '\n') { |
30944b6d IZ |
968 | /* Eventually fbm_*() should handle this, but often |
969 | anchored_offset is not 0, so this check will not be wasted. */ | |
970 | /* XXXX In the code below we prefer to look for "^" even in | |
971 | presence of anchored substrings. And we search even | |
972 | beyond the found float position. These pessimizations | |
973 | are historical artefacts only. */ | |
974 | find_anchor: | |
2c2d71f5 | 975 | while (t < strend - prog->minlen) { |
cad2e5aa | 976 | if (*t == '\n') { |
4ee3650e | 977 | if (t < check_at - prog->check_offset_min) { |
f2ed9b32 | 978 | if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) { |
4ee3650e GS |
979 | /* Since we moved from the found position, |
980 | we definitely contradict the found anchored | |
30944b6d IZ |
981 | substr. Due to the above check we do not |
982 | contradict "check" substr. | |
983 | Thus we can arrive here only if check substr | |
984 | is float. Redo checking for "other"=="fixed". | |
985 | */ | |
9041c2e3 | 986 | strpos = t + 1; |
a3621e74 | 987 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n", |
e4584336 | 988 | PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); |
30944b6d IZ |
989 | goto do_other_anchored; |
990 | } | |
4ee3650e GS |
991 | /* We don't contradict the found floating substring. */ |
992 | /* XXXX Why not check for STCLASS? */ | |
cad2e5aa | 993 | s = t + 1; |
a3621e74 | 994 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", |
e4584336 | 995 | PL_colors[0], PL_colors[1], (long)(s - i_strpos))); |
cad2e5aa JH |
996 | goto set_useful; |
997 | } | |
4ee3650e GS |
998 | /* Position contradicts check-string */ |
999 | /* XXXX probably better to look for check-string | |
1000 | than for "\n", so one should lower the limit for t? */ | |
a3621e74 | 1001 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", |
e4584336 | 1002 | PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos))); |
0e41cd87 | 1003 | other_last = strpos = s = t + 1; |
cad2e5aa JH |
1004 | goto restart; |
1005 | } | |
1006 | t++; | |
1007 | } | |
a3621e74 | 1008 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n", |
e4584336 | 1009 | PL_colors[0], PL_colors[1])); |
2c2d71f5 | 1010 | goto fail_finish; |
cad2e5aa | 1011 | } |
f5952150 | 1012 | else { |
a3621e74 | 1013 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", |
e4584336 | 1014 | PL_colors[0], PL_colors[1])); |
f5952150 | 1015 | } |
cad2e5aa JH |
1016 | s = t; |
1017 | set_useful: | |
f2ed9b32 | 1018 | ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */ |
cad2e5aa JH |
1019 | } |
1020 | else { | |
f5952150 | 1021 | /* The found string does not prohibit matching at strpos, |
2c2d71f5 | 1022 | - no optimization of calling REx engine can be performed, |
f5952150 GS |
1023 | unless it was an MBOL and we are not after MBOL, |
1024 | or a future STCLASS check will fail this. */ | |
2c2d71f5 JH |
1025 | try_at_start: |
1026 | /* Even in this situation we may use MBOL flag if strpos is offset | |
1027 | wrt the start of the string. */ | |
05b4157f | 1028 | if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */ |
a1933d95 | 1029 | && (strpos != strbeg) && strpos[-1] != '\n' |
d506a20d | 1030 | /* May be due to an implicit anchor of m{.*foo} */ |
bbe252da | 1031 | && !(prog->intflags & PREGf_IMPLICIT)) |
d506a20d | 1032 | { |
cad2e5aa JH |
1033 | t = strpos; |
1034 | goto find_anchor; | |
1035 | } | |
a3621e74 | 1036 | DEBUG_EXECUTE_r( if (ml_anch) |
f5952150 | 1037 | PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", |
70685ca0 | 1038 | (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); |
30944b6d | 1039 | ); |
2c2d71f5 | 1040 | success_at_start: |
bbe252da | 1041 | if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */ |
f2ed9b32 | 1042 | && (utf8_target ? ( |
33b8afdf JH |
1043 | prog->check_utf8 /* Could be deleted already */ |
1044 | && --BmUSEFUL(prog->check_utf8) < 0 | |
1045 | && (prog->check_utf8 == prog->float_utf8) | |
1046 | ) : ( | |
1047 | prog->check_substr /* Could be deleted already */ | |
1048 | && --BmUSEFUL(prog->check_substr) < 0 | |
1049 | && (prog->check_substr == prog->float_substr) | |
1050 | ))) | |
66e933ab | 1051 | { |
cad2e5aa | 1052 | /* If flags & SOMETHING - do not do it many times on the same match */ |
a3621e74 | 1053 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n")); |
f2ed9b32 KW |
1054 | /* XXX Does the destruction order has to change with utf8_target? */ |
1055 | SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr); | |
1056 | SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8); | |
a0714e2c SS |
1057 | prog->check_substr = prog->check_utf8 = NULL; /* disable */ |
1058 | prog->float_substr = prog->float_utf8 = NULL; /* clear */ | |
1059 | check = NULL; /* abort */ | |
cad2e5aa | 1060 | s = strpos; |
486ec47a | 1061 | /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevant flag |
c9415951 YO |
1062 | see http://bugs.activestate.com/show_bug.cgi?id=87173 */ |
1063 | if (prog->intflags & PREGf_IMPLICIT) | |
1064 | prog->extflags &= ~RXf_ANCH_MBOL; | |
3cf5c195 IZ |
1065 | /* XXXX This is a remnant of the old implementation. It |
1066 | looks wasteful, since now INTUIT can use many | |
6eb5f6b9 | 1067 | other heuristics. */ |
bbe252da | 1068 | prog->extflags &= ~RXf_USE_INTUIT; |
c9415951 | 1069 | /* XXXX What other flags might need to be cleared in this branch? */ |
cad2e5aa JH |
1070 | } |
1071 | else | |
1072 | s = strpos; | |
1073 | } | |
1074 | ||
6eb5f6b9 JH |
1075 | /* Last resort... */ |
1076 | /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ | |
1de06328 YO |
1077 | /* trie stclasses are too expensive to use here, we are better off to |
1078 | leave it to regmatch itself */ | |
f8fc2ecf | 1079 | if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) { |
6eb5f6b9 JH |
1080 | /* minlen == 0 is possible if regstclass is \b or \B, |
1081 | and the fixed substr is ''$. | |
1082 | Since minlen is already taken into account, s+1 is before strend; | |
1083 | accidentally, minlen >= 1 guaranties no false positives at s + 1 | |
1084 | even for \b or \B. But (minlen? 1 : 0) below assumes that | |
1085 | regstclass does not come from lookahead... */ | |
1086 | /* If regstclass takes bytelength more than 1: If charlength==1, OK. | |
af944926 | 1087 | This leaves EXACTF-ish only, which are dealt with in find_byclass(). */ |
f8fc2ecf YO |
1088 | const U8* const str = (U8*)STRING(progi->regstclass); |
1089 | const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT | |
1090 | ? CHR_DIST(str+STR_LEN(progi->regstclass), str) | |
66e933ab | 1091 | : 1); |
1de06328 YO |
1092 | char * endpos; |
1093 | if (prog->anchored_substr || prog->anchored_utf8 || ml_anch) | |
1094 | endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend); | |
1095 | else if (prog->float_substr || prog->float_utf8) | |
1096 | endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend); | |
1097 | else | |
1098 | endpos= strend; | |
1099 | ||
d8080198 YO |
1100 | if (checked_upto < s) |
1101 | checked_upto = s; | |
1102 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n", | |
1103 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg))); | |
1104 | ||
6eb5f6b9 | 1105 | t = s; |
d8080198 YO |
1106 | s = find_byclass(prog, progi->regstclass, checked_upto, endpos, NULL); |
1107 | if (s) { | |
1108 | checked_upto = s; | |
1109 | } else { | |
6eb5f6b9 | 1110 | #ifdef DEBUGGING |
cbbf8932 | 1111 | const char *what = NULL; |
6eb5f6b9 JH |
1112 | #endif |
1113 | if (endpos == strend) { | |
a3621e74 | 1114 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
6eb5f6b9 JH |
1115 | "Could not match STCLASS...\n") ); |
1116 | goto fail; | |
1117 | } | |
a3621e74 | 1118 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab | 1119 | "This position contradicts STCLASS...\n") ); |
bbe252da | 1120 | if ((prog->extflags & RXf_ANCH) && !ml_anch) |
653099ff | 1121 | goto fail; |
d8080198 YO |
1122 | checked_upto = HOPBACKc(endpos, start_shift); |
1123 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n", | |
1124 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg))); | |
6eb5f6b9 | 1125 | /* Contradict one of substrings */ |
33b8afdf | 1126 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 | 1127 | if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) { |
a3621e74 | 1128 | DEBUG_EXECUTE_r( what = "anchored" ); |
6eb5f6b9 | 1129 | hop_and_restart: |
1aa99e6b | 1130 | s = HOP3c(t, 1, strend); |
66e933ab GS |
1131 | if (s + start_shift + end_shift > strend) { |
1132 | /* XXXX Should be taken into account earlier? */ | |
a3621e74 | 1133 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab GS |
1134 | "Could not match STCLASS...\n") ); |
1135 | goto fail; | |
1136 | } | |
5e39e1e5 HS |
1137 | if (!check) |
1138 | goto giveup; | |
a3621e74 | 1139 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1140 | "Looking for %s substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1141 | what, (long)(s + start_shift - i_strpos)) ); |
1142 | goto restart; | |
1143 | } | |
66e933ab | 1144 | /* Have both, check_string is floating */ |
6eb5f6b9 JH |
1145 | if (t + start_shift >= check_at) /* Contradicts floating=check */ |
1146 | goto retry_floating_check; | |
1147 | /* Recheck anchored substring, but not floating... */ | |
9041c2e3 | 1148 | s = check_at; |
5e39e1e5 HS |
1149 | if (!check) |
1150 | goto giveup; | |
a3621e74 | 1151 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1152 | "Looking for anchored substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1153 | (long)(other_last - i_strpos)) ); |
1154 | goto do_other_anchored; | |
1155 | } | |
60e71179 GS |
1156 | /* Another way we could have checked stclass at the |
1157 | current position only: */ | |
1158 | if (ml_anch) { | |
1159 | s = t = t + 1; | |
5e39e1e5 HS |
1160 | if (!check) |
1161 | goto giveup; | |
a3621e74 | 1162 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1163 | "Looking for /%s^%s/m starting at offset %ld...\n", |
e4584336 | 1164 | PL_colors[0], PL_colors[1], (long)(t - i_strpos)) ); |
60e71179 | 1165 | goto try_at_offset; |
66e933ab | 1166 | } |
f2ed9b32 | 1167 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */ |
60e71179 | 1168 | goto fail; |
486ec47a | 1169 | /* Check is floating substring. */ |
6eb5f6b9 JH |
1170 | retry_floating_check: |
1171 | t = check_at - start_shift; | |
a3621e74 | 1172 | DEBUG_EXECUTE_r( what = "floating" ); |
6eb5f6b9 JH |
1173 | goto hop_and_restart; |
1174 | } | |
b7953727 | 1175 | if (t != s) { |
a3621e74 | 1176 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
6eb5f6b9 | 1177 | "By STCLASS: moving %ld --> %ld\n", |
b7953727 JH |
1178 | (long)(t - i_strpos), (long)(s - i_strpos)) |
1179 | ); | |
1180 | } | |
1181 | else { | |
a3621e74 | 1182 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b7953727 JH |
1183 | "Does not contradict STCLASS...\n"); |
1184 | ); | |
1185 | } | |
6eb5f6b9 | 1186 | } |
5e39e1e5 | 1187 | giveup: |
a3621e74 | 1188 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n", |
5e39e1e5 HS |
1189 | PL_colors[4], (check ? "Guessed" : "Giving up"), |
1190 | PL_colors[5], (long)(s - i_strpos)) ); | |
cad2e5aa | 1191 | return s; |
2c2d71f5 JH |
1192 | |
1193 | fail_finish: /* Substring not found */ | |
33b8afdf | 1194 | if (prog->check_substr || prog->check_utf8) /* could be removed already */ |
f2ed9b32 | 1195 | BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */ |
cad2e5aa | 1196 | fail: |
a3621e74 | 1197 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", |
e4584336 | 1198 | PL_colors[4], PL_colors[5])); |
bd61b366 | 1199 | return NULL; |
cad2e5aa | 1200 | } |
9661b544 | 1201 | |
a0a388a1 YO |
1202 | #define DECL_TRIE_TYPE(scan) \ |
1203 | const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \ | |
fab2782b YO |
1204 | trie_type = ((scan->flags == EXACT) \ |
1205 | ? (utf8_target ? trie_utf8 : trie_plain) \ | |
1206 | : (utf8_target ? trie_utf8_fold : trie_latin_utf8_fold)) | |
1207 | ||
1208 | #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \ | |
1209 | uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ | |
1210 | STRLEN skiplen; \ | |
1211 | switch (trie_type) { \ | |
1212 | case trie_utf8_fold: \ | |
1213 | if ( foldlen>0 ) { \ | |
1214 | uvc = utf8n_to_uvuni( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1215 | foldlen -= len; \ | |
1216 | uscan += len; \ | |
1217 | len=0; \ | |
1218 | } else { \ | |
1219 | uvc = to_utf8_fold( (const U8*) uc, foldbuf, &foldlen ); \ | |
1220 | len = UTF8SKIP(uc); \ | |
1221 | skiplen = UNISKIP( uvc ); \ | |
1222 | foldlen -= skiplen; \ | |
1223 | uscan = foldbuf + skiplen; \ | |
1224 | } \ | |
1225 | break; \ | |
1226 | case trie_latin_utf8_fold: \ | |
1227 | if ( foldlen>0 ) { \ | |
1228 | uvc = utf8n_to_uvuni( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1229 | foldlen -= len; \ | |
1230 | uscan += len; \ | |
1231 | len=0; \ | |
1232 | } else { \ | |
1233 | len = 1; \ | |
1234 | uvc = _to_fold_latin1( (U8) *uc, foldbuf, &foldlen, 1); \ | |
1235 | skiplen = UNISKIP( uvc ); \ | |
1236 | foldlen -= skiplen; \ | |
1237 | uscan = foldbuf + skiplen; \ | |
1238 | } \ | |
1239 | break; \ | |
1240 | case trie_utf8: \ | |
1241 | uvc = utf8n_to_uvuni( (const U8*) uc, UTF8_MAXLEN, &len, uniflags ); \ | |
1242 | break; \ | |
1243 | case trie_plain: \ | |
1244 | uvc = (UV)*uc; \ | |
1245 | len = 1; \ | |
1246 | } \ | |
1247 | if (uvc < 256) { \ | |
1248 | charid = trie->charmap[ uvc ]; \ | |
1249 | } \ | |
1250 | else { \ | |
1251 | charid = 0; \ | |
1252 | if (widecharmap) { \ | |
1253 | SV** const svpp = hv_fetch(widecharmap, \ | |
1254 | (char*)&uvc, sizeof(UV), 0); \ | |
1255 | if (svpp) \ | |
1256 | charid = (U16)SvIV(*svpp); \ | |
1257 | } \ | |
1258 | } \ | |
4cadc6a9 YO |
1259 | } STMT_END |
1260 | ||
4cadc6a9 YO |
1261 | #define REXEC_FBC_EXACTISH_SCAN(CoNd) \ |
1262 | STMT_START { \ | |
1263 | while (s <= e) { \ | |
1264 | if ( (CoNd) \ | |
fac1af77 | 1265 | && (ln == 1 || folder(s, pat_string, ln)) \ |
9a5a5549 | 1266 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1267 | goto got_it; \ |
1268 | s++; \ | |
1269 | } \ | |
1270 | } STMT_END | |
1271 | ||
1272 | #define REXEC_FBC_UTF8_SCAN(CoDe) \ | |
1273 | STMT_START { \ | |
7016d6eb | 1274 | while (s < strend && s + (uskip = UTF8SKIP(s)) <= strend) { \ |
4cadc6a9 YO |
1275 | CoDe \ |
1276 | s += uskip; \ | |
1277 | } \ | |
1278 | } STMT_END | |
1279 | ||
1280 | #define REXEC_FBC_SCAN(CoDe) \ | |
1281 | STMT_START { \ | |
1282 | while (s < strend) { \ | |
1283 | CoDe \ | |
1284 | s++; \ | |
1285 | } \ | |
1286 | } STMT_END | |
1287 | ||
1288 | #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ | |
1289 | REXEC_FBC_UTF8_SCAN( \ | |
1290 | if (CoNd) { \ | |
24b23f37 | 1291 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1292 | goto got_it; \ |
1293 | else \ | |
1294 | tmp = doevery; \ | |
1295 | } \ | |
1296 | else \ | |
1297 | tmp = 1; \ | |
1298 | ) | |
1299 | ||
1300 | #define REXEC_FBC_CLASS_SCAN(CoNd) \ | |
1301 | REXEC_FBC_SCAN( \ | |
1302 | if (CoNd) { \ | |
24b23f37 | 1303 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1304 | goto got_it; \ |
1305 | else \ | |
1306 | tmp = doevery; \ | |
1307 | } \ | |
1308 | else \ | |
1309 | tmp = 1; \ | |
1310 | ) | |
1311 | ||
1312 | #define REXEC_FBC_TRYIT \ | |
24b23f37 | 1313 | if ((!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1314 | goto got_it |
1315 | ||
e1d1eefb | 1316 | #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \ |
f2ed9b32 | 1317 | if (utf8_target) { \ |
e1d1eefb YO |
1318 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1319 | } \ | |
1320 | else { \ | |
1321 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1322 | } |
e1d1eefb | 1323 | |
4cadc6a9 | 1324 | #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \ |
f2ed9b32 | 1325 | if (utf8_target) { \ |
4cadc6a9 YO |
1326 | UtFpReLoAd; \ |
1327 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ | |
1328 | } \ | |
1329 | else { \ | |
1330 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1331 | } |
4cadc6a9 YO |
1332 | |
1333 | #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \ | |
1334 | PL_reg_flags |= RF_tainted; \ | |
f2ed9b32 | 1335 | if (utf8_target) { \ |
4cadc6a9 YO |
1336 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1337 | } \ | |
1338 | else { \ | |
1339 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1340 | } |
4cadc6a9 | 1341 | |
786e8c11 YO |
1342 | #define DUMP_EXEC_POS(li,s,doutf8) \ |
1343 | dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8) | |
1344 | ||
cfaf538b KW |
1345 | |
1346 | #define UTF8_NOLOAD(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ | |
1347 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1348 | tmp = TEST_NON_UTF8(tmp); \ | |
1349 | REXEC_FBC_UTF8_SCAN( \ | |
1350 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1351 | tmp = !tmp; \ | |
1352 | IF_SUCCESS; \ | |
1353 | } \ | |
1354 | else { \ | |
1355 | IF_FAIL; \ | |
1356 | } \ | |
1357 | ); \ | |
1358 | ||
1359 | #define UTF8_LOAD(TeSt1_UtF8, TeSt2_UtF8, IF_SUCCESS, IF_FAIL) \ | |
1360 | if (s == PL_bostr) { \ | |
1361 | tmp = '\n'; \ | |
1362 | } \ | |
1363 | else { \ | |
1364 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); \ | |
1365 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); \ | |
1366 | } \ | |
1367 | tmp = TeSt1_UtF8; \ | |
1368 | LOAD_UTF8_CHARCLASS_ALNUM(); \ | |
1369 | REXEC_FBC_UTF8_SCAN( \ | |
1370 | if (tmp == ! (TeSt2_UtF8)) { \ | |
1371 | tmp = !tmp; \ | |
1372 | IF_SUCCESS; \ | |
1373 | } \ | |
1374 | else { \ | |
1375 | IF_FAIL; \ | |
1376 | } \ | |
1377 | ); \ | |
1378 | ||
63ac0dad KW |
1379 | /* The only difference between the BOUND and NBOUND cases is that |
1380 | * REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in | |
1381 | * NBOUND. This is accomplished by passing it in either the if or else clause, | |
1382 | * with the other one being empty */ | |
1383 | #define FBC_BOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1384 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
cfaf538b KW |
1385 | |
1386 | #define FBC_BOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1387 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
63ac0dad KW |
1388 | |
1389 | #define FBC_NBOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1390 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b KW |
1391 | |
1392 | #define FBC_NBOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1393 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b | 1394 | |
63ac0dad KW |
1395 | |
1396 | /* Common to the BOUND and NBOUND cases. Unfortunately the UTF8 tests need to | |
1397 | * be passed in completely with the variable name being tested, which isn't | |
1398 | * such a clean interface, but this is easier to read than it was before. We | |
1399 | * are looking for the boundary (or non-boundary between a word and non-word | |
1400 | * character. The utf8 and non-utf8 cases have the same logic, but the details | |
1401 | * must be different. Find the "wordness" of the character just prior to this | |
1402 | * one, and compare it with the wordness of this one. If they differ, we have | |
1403 | * a boundary. At the beginning of the string, pretend that the previous | |
1404 | * character was a new-line */ | |
cfaf538b | 1405 | #define FBC_BOUND_COMMON(UTF8_CODE, TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ |
63ac0dad | 1406 | if (utf8_target) { \ |
cfaf538b | 1407 | UTF8_CODE \ |
63ac0dad KW |
1408 | } \ |
1409 | else { /* Not utf8 */ \ | |
1410 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1411 | tmp = TEST_NON_UTF8(tmp); \ | |
1412 | REXEC_FBC_SCAN( \ | |
1413 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1414 | tmp = !tmp; \ | |
1415 | IF_SUCCESS; \ | |
1416 | } \ | |
1417 | else { \ | |
1418 | IF_FAIL; \ | |
1419 | } \ | |
1420 | ); \ | |
1421 | } \ | |
1422 | if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s))) \ | |
1423 | goto got_it; | |
1424 | ||
786e8c11 YO |
1425 | /* We know what class REx starts with. Try to find this position... */ |
1426 | /* if reginfo is NULL, its a dryrun */ | |
1427 | /* annoyingly all the vars in this routine have different names from their counterparts | |
1428 | in regmatch. /grrr */ | |
1429 | ||
3c3eec57 | 1430 | STATIC char * |
07be1b83 | 1431 | S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, |
24b23f37 | 1432 | const char *strend, regmatch_info *reginfo) |
a687059c | 1433 | { |
27da23d5 | 1434 | dVAR; |
bbe252da | 1435 | const I32 doevery = (prog->intflags & PREGf_SKIP) == 0; |
fac1af77 KW |
1436 | char *pat_string; /* The pattern's exactish string */ |
1437 | char *pat_end; /* ptr to end char of pat_string */ | |
1438 | re_fold_t folder; /* Function for computing non-utf8 folds */ | |
1439 | const U8 *fold_array; /* array for folding ords < 256 */ | |
d8093b23 | 1440 | STRLEN ln; |
5dab1207 | 1441 | STRLEN lnc; |
eb578fdb | 1442 | STRLEN uskip; |
fac1af77 KW |
1443 | U8 c1; |
1444 | U8 c2; | |
6eb5f6b9 | 1445 | char *e; |
eb578fdb KW |
1446 | I32 tmp = 1; /* Scratch variable? */ |
1447 | const bool utf8_target = PL_reg_match_utf8; | |
453bfd44 | 1448 | UV utf8_fold_flags = 0; |
f8fc2ecf | 1449 | RXi_GET_DECL(prog,progi); |
7918f24d NC |
1450 | |
1451 | PERL_ARGS_ASSERT_FIND_BYCLASS; | |
f8fc2ecf | 1452 | |
6eb5f6b9 JH |
1453 | /* We know what class it must start with. */ |
1454 | switch (OP(c)) { | |
6eb5f6b9 | 1455 | case ANYOF: |
e0193e47 | 1456 | if (utf8_target) { |
b1e3e569 KW |
1457 | STRLEN inclasslen = strend - s; |
1458 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1459 | reginclass(prog, c, (U8*)s, &inclasslen, utf8_target)); | |
388cc4de HS |
1460 | } |
1461 | else { | |
6ef69d56 | 1462 | REXEC_FBC_CLASS_SCAN(REGINCLASS(prog, c, (U8*)s)); |
a0d0e21e | 1463 | } |
6eb5f6b9 | 1464 | break; |
f33976b4 | 1465 | case CANY: |
4cadc6a9 | 1466 | REXEC_FBC_SCAN( |
24b23f37 | 1467 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
f33976b4 DB |
1468 | goto got_it; |
1469 | else | |
1470 | tmp = doevery; | |
4cadc6a9 | 1471 | ); |
f33976b4 | 1472 | break; |
2f7f8cb1 KW |
1473 | |
1474 | case EXACTFA: | |
1475 | if (UTF_PATTERN || utf8_target) { | |
1476 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
1477 | goto do_exactf_utf8; | |
1478 | } | |
1479 | fold_array = PL_fold_latin1; /* Latin1 folds are not affected by */ | |
1480 | folder = foldEQ_latin1; /* /a, except the sharp s one which */ | |
1481 | goto do_exactf_non_utf8; /* isn't dealt with by these */ | |
1482 | ||
6eb5f6b9 | 1483 | case EXACTF: |
62bf7766 | 1484 | if (utf8_target) { |
77a6d856 KW |
1485 | |
1486 | /* regcomp.c already folded this if pattern is in UTF-8 */ | |
62bf7766 | 1487 | utf8_fold_flags = 0; |
fac1af77 KW |
1488 | goto do_exactf_utf8; |
1489 | } | |
1490 | fold_array = PL_fold; | |
1491 | folder = foldEQ; | |
1492 | goto do_exactf_non_utf8; | |
1493 | ||
1494 | case EXACTFL: | |
1495 | if (UTF_PATTERN || utf8_target) { | |
17580e7a | 1496 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
fac1af77 KW |
1497 | goto do_exactf_utf8; |
1498 | } | |
1499 | fold_array = PL_fold_locale; | |
1500 | folder = foldEQ_locale; | |
16d951b7 KW |
1501 | goto do_exactf_non_utf8; |
1502 | ||
3c760661 KW |
1503 | case EXACTFU_SS: |
1504 | if (UTF_PATTERN) { | |
1505 | utf8_fold_flags = FOLDEQ_S2_ALREADY_FOLDED; | |
1506 | } | |
1507 | goto do_exactf_utf8; | |
1508 | ||
fab2782b | 1509 | case EXACTFU_TRICKYFOLD: |
16d951b7 KW |
1510 | case EXACTFU: |
1511 | if (UTF_PATTERN || utf8_target) { | |
77a6d856 | 1512 | utf8_fold_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0; |
16d951b7 KW |
1513 | goto do_exactf_utf8; |
1514 | } | |
1515 | ||
1516 | /* Any 'ss' in the pattern should have been replaced by regcomp, | |
1517 | * so we don't have to worry here about this single special case | |
1518 | * in the Latin1 range */ | |
1519 | fold_array = PL_fold_latin1; | |
1520 | folder = foldEQ_latin1; | |
fac1af77 KW |
1521 | |
1522 | /* FALL THROUGH */ | |
1523 | ||
62bf7766 KW |
1524 | do_exactf_non_utf8: /* Neither pattern nor string are UTF8, and there |
1525 | are no glitches with fold-length differences | |
1526 | between the target string and pattern */ | |
fac1af77 KW |
1527 | |
1528 | /* The idea in the non-utf8 EXACTF* cases is to first find the | |
1529 | * first character of the EXACTF* node and then, if necessary, | |
1530 | * case-insensitively compare the full text of the node. c1 is the | |
1531 | * first character. c2 is its fold. This logic will not work for | |
1532 | * Unicode semantics and the german sharp ss, which hence should | |
1533 | * not be compiled into a node that gets here. */ | |
1534 | pat_string = STRING(c); | |
1535 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1536 | ||
8a90a8fe KW |
1537 | /* We know that we have to match at least 'ln' bytes (which is the |
1538 | * same as characters, since not utf8). If we have to match 3 | |
1539 | * characters, and there are only 2 availabe, we know without | |
1540 | * trying that it will fail; so don't start a match past the | |
1541 | * required minimum number from the far end */ | |
fac1af77 KW |
1542 | e = HOP3c(strend, -((I32)ln), s); |
1543 | ||
1544 | if (!reginfo && e < s) { | |
1545 | e = s; /* Due to minlen logic of intuit() */ | |
1546 | } | |
1547 | ||
1548 | c1 = *pat_string; | |
1549 | c2 = fold_array[c1]; | |
1550 | if (c1 == c2) { /* If char and fold are the same */ | |
1551 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1); | |
1552 | } | |
1553 | else { | |
1554 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2); | |
1555 | } | |
1556 | break; | |
1557 | ||
1558 | do_exactf_utf8: | |
e067297c KW |
1559 | { |
1560 | unsigned expansion; | |
1561 | ||
fac1af77 KW |
1562 | |
1563 | /* If one of the operands is in utf8, we can't use the simpler | |
1564 | * folding above, due to the fact that many different characters | |
1565 | * can have the same fold, or portion of a fold, or different- | |
1566 | * length fold */ | |
1567 | pat_string = STRING(c); | |
1568 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1569 | pat_end = pat_string + ln; | |
1570 | lnc = (UTF_PATTERN) /* length to match in characters */ | |
1571 | ? utf8_length((U8 *) pat_string, (U8 *) pat_end) | |
1572 | : ln; | |
1573 | ||
e067297c KW |
1574 | /* We have 'lnc' characters to match in the pattern, but because of |
1575 | * multi-character folding, each character in the target can match | |
1576 | * up to 3 characters (Unicode guarantees it will never exceed | |
1577 | * this) if it is utf8-encoded; and up to 2 if not (based on the | |
1578 | * fact that the Latin 1 folds are already determined, and the | |
1579 | * only multi-char fold in that range is the sharp-s folding to | |
1580 | * 'ss'. Thus, a pattern character can match as little as 1/3 of a | |
89378d8a KW |
1581 | * string character. Adjust lnc accordingly, rounding up, so that |
1582 | * if we need to match at least 4+1/3 chars, that really is 5. */ | |
e067297c | 1583 | expansion = (utf8_target) ? UTF8_MAX_FOLD_CHAR_EXPAND : 2; |
89378d8a | 1584 | lnc = (lnc + expansion - 1) / expansion; |
e067297c KW |
1585 | |
1586 | /* As in the non-UTF8 case, if we have to match 3 characters, and | |
1587 | * only 2 are left, it's guaranteed to fail, so don't start a | |
1588 | * match that would require us to go beyond the end of the string | |
1589 | */ | |
1590 | e = HOP3c(strend, -((I32)lnc), s); | |
fac1af77 KW |
1591 | |
1592 | if (!reginfo && e < s) { | |
1593 | e = s; /* Due to minlen logic of intuit() */ | |
1594 | } | |
1595 | ||
b33105db KW |
1596 | /* XXX Note that we could recalculate e to stop the loop earlier, |
1597 | * as the worst case expansion above will rarely be met, and as we | |
1598 | * go along we would usually find that e moves further to the left. | |
1599 | * This would happen only after we reached the point in the loop | |
1600 | * where if there were no expansion we should fail. Unclear if | |
1601 | * worth the expense */ | |
e067297c | 1602 | |
fac1af77 KW |
1603 | while (s <= e) { |
1604 | char *my_strend= (char *)strend; | |
1605 | if (foldEQ_utf8_flags(s, &my_strend, 0, utf8_target, | |
1606 | pat_string, NULL, ln, cBOOL(UTF_PATTERN), utf8_fold_flags) | |
1607 | && (!reginfo || regtry(reginfo, &s)) ) | |
1608 | { | |
1609 | goto got_it; | |
1610 | } | |
bbdd8bad | 1611 | s += (utf8_target) ? UTF8SKIP(s) : 1; |
fac1af77 KW |
1612 | } |
1613 | break; | |
e067297c | 1614 | } |
bbce6d69 | 1615 | case BOUNDL: |
3280af22 | 1616 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1617 | FBC_BOUND(isALNUM_LC, |
1618 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1619 | isALNUM_LC_utf8((U8*)s)); | |
a0ed51b3 | 1620 | break; |
bbce6d69 | 1621 | case NBOUNDL: |
3280af22 | 1622 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1623 | FBC_NBOUND(isALNUM_LC, |
1624 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1625 | isALNUM_LC_utf8((U8*)s)); | |
1626 | break; | |
1627 | case BOUND: | |
1628 | FBC_BOUND(isWORDCHAR, | |
1629 | isALNUM_uni(tmp), | |
1630 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1631 | break; | |
cfaf538b KW |
1632 | case BOUNDA: |
1633 | FBC_BOUND_NOLOAD(isWORDCHAR_A, | |
1634 | isWORDCHAR_A(tmp), | |
1635 | isWORDCHAR_A((U8*)s)); | |
1636 | break; | |
a0d0e21e | 1637 | case NBOUND: |
63ac0dad KW |
1638 | FBC_NBOUND(isWORDCHAR, |
1639 | isALNUM_uni(tmp), | |
1640 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1641 | break; | |
cfaf538b KW |
1642 | case NBOUNDA: |
1643 | FBC_NBOUND_NOLOAD(isWORDCHAR_A, | |
1644 | isWORDCHAR_A(tmp), | |
1645 | isWORDCHAR_A((U8*)s)); | |
1646 | break; | |
63ac0dad KW |
1647 | case BOUNDU: |
1648 | FBC_BOUND(isWORDCHAR_L1, | |
1649 | isALNUM_uni(tmp), | |
1650 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1651 | break; | |
1652 | case NBOUNDU: | |
1653 | FBC_NBOUND(isWORDCHAR_L1, | |
1654 | isALNUM_uni(tmp), | |
1655 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
a0ed51b3 | 1656 | break; |
bbce6d69 | 1657 | case ALNUML: |
4cadc6a9 YO |
1658 | REXEC_FBC_CSCAN_TAINT( |
1659 | isALNUM_LC_utf8((U8*)s), | |
1660 | isALNUM_LC(*s) | |
1661 | ); | |
6895a8aa | 1662 | break; |
980866de KW |
1663 | case ALNUMU: |
1664 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1665 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1666 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1667 | isWORDCHAR_L1((U8) *s) |
1668 | ); | |
6895a8aa | 1669 | break; |
980866de KW |
1670 | case ALNUM: |
1671 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1672 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1673 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1674 | isWORDCHAR((U8) *s) |
1675 | ); | |
6895a8aa | 1676 | break; |
cfaf538b | 1677 | case ALNUMA: |
8e9da4d4 KW |
1678 | /* Don't need to worry about utf8, as it can match only a single |
1679 | * byte invariant character */ | |
cfaf538b | 1680 | REXEC_FBC_CLASS_SCAN( isWORDCHAR_A(*s)); |
6895a8aa | 1681 | break; |
980866de KW |
1682 | case NALNUMU: |
1683 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 | 1684 | LOAD_UTF8_CHARCLASS_ALNUM(), |
359960d4 | 1685 | !swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), |
980866de KW |
1686 | ! isWORDCHAR_L1((U8) *s) |
1687 | ); | |
6895a8aa | 1688 | break; |
a0d0e21e | 1689 | case NALNUM: |
4cadc6a9 | 1690 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1691 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1692 | !swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target), | |
980866de | 1693 | ! isALNUM(*s) |
4cadc6a9 | 1694 | ); |
6895a8aa | 1695 | break; |
cfaf538b | 1696 | case NALNUMA: |
8e9da4d4 KW |
1697 | REXEC_FBC_CSCAN( |
1698 | !isWORDCHAR_A(*s), | |
1699 | !isWORDCHAR_A(*s) | |
1700 | ); | |
1701 | break; | |
bbce6d69 | 1702 | case NALNUML: |
4cadc6a9 YO |
1703 | REXEC_FBC_CSCAN_TAINT( |
1704 | !isALNUM_LC_utf8((U8*)s), | |
1705 | !isALNUM_LC(*s) | |
1706 | ); | |
6895a8aa | 1707 | break; |
980866de KW |
1708 | case SPACEU: |
1709 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1710 | LOAD_UTF8_CHARCLASS_SPACE(), |
1711 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de KW |
1712 | isSPACE_L1((U8) *s) |
1713 | ); | |
6895a8aa | 1714 | break; |
a0d0e21e | 1715 | case SPACE: |
4cadc6a9 | 1716 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1717 | LOAD_UTF8_CHARCLASS_SPACE(), |
1718 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de | 1719 | isSPACE((U8) *s) |
4cadc6a9 | 1720 | ); |
6895a8aa | 1721 | break; |
cfaf538b | 1722 | case SPACEA: |
8e9da4d4 KW |
1723 | /* Don't need to worry about utf8, as it can match only a single |
1724 | * byte invariant character */ | |
cfaf538b | 1725 | REXEC_FBC_CLASS_SCAN( isSPACE_A(*s)); |
6895a8aa | 1726 | break; |
bbce6d69 | 1727 | case SPACEL: |
4cadc6a9 | 1728 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1729 | isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1730 | isSPACE_LC(*s) |
1731 | ); | |
6895a8aa | 1732 | break; |
980866de KW |
1733 | case NSPACEU: |
1734 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1735 | LOAD_UTF8_CHARCLASS_SPACE(), |
1736 | !( *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de KW |
1737 | ! isSPACE_L1((U8) *s) |
1738 | ); | |
6895a8aa | 1739 | break; |
a0d0e21e | 1740 | case NSPACE: |
4cadc6a9 | 1741 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1742 | LOAD_UTF8_CHARCLASS_SPACE(), |
1743 | !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de | 1744 | ! isSPACE((U8) *s) |
4cadc6a9 | 1745 | ); |
6895a8aa | 1746 | break; |
cfaf538b | 1747 | case NSPACEA: |
8e9da4d4 KW |
1748 | REXEC_FBC_CSCAN( |
1749 | !isSPACE_A(*s), | |
1750 | !isSPACE_A(*s) | |
1751 | ); | |
1752 | break; | |
bbce6d69 | 1753 | case NSPACEL: |
4cadc6a9 | 1754 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1755 | !isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1756 | !isSPACE_LC(*s) |
1757 | ); | |
6895a8aa | 1758 | break; |
a0d0e21e | 1759 | case DIGIT: |
4cadc6a9 | 1760 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1761 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1762 | swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1763 | isDIGIT(*s) |
1764 | ); | |
6895a8aa | 1765 | break; |
cfaf538b | 1766 | case DIGITA: |
8e9da4d4 KW |
1767 | /* Don't need to worry about utf8, as it can match only a single |
1768 | * byte invariant character */ | |
cfaf538b | 1769 | REXEC_FBC_CLASS_SCAN( isDIGIT_A(*s)); |
6895a8aa | 1770 | break; |
b8c5462f | 1771 | case DIGITL: |
4cadc6a9 YO |
1772 | REXEC_FBC_CSCAN_TAINT( |
1773 | isDIGIT_LC_utf8((U8*)s), | |
1774 | isDIGIT_LC(*s) | |
1775 | ); | |
6895a8aa | 1776 | break; |
a0d0e21e | 1777 | case NDIGIT: |
4cadc6a9 | 1778 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1779 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1780 | !swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1781 | !isDIGIT(*s) |
1782 | ); | |
6895a8aa | 1783 | break; |
cfaf538b | 1784 | case NDIGITA: |
8e9da4d4 KW |
1785 | REXEC_FBC_CSCAN( |
1786 | !isDIGIT_A(*s), | |
1787 | !isDIGIT_A(*s) | |
1788 | ); | |
1789 | break; | |
b8c5462f | 1790 | case NDIGITL: |
4cadc6a9 YO |
1791 | REXEC_FBC_CSCAN_TAINT( |
1792 | !isDIGIT_LC_utf8((U8*)s), | |
1793 | !isDIGIT_LC(*s) | |
1794 | ); | |
6895a8aa | 1795 | break; |
e1d1eefb YO |
1796 | case LNBREAK: |
1797 | REXEC_FBC_CSCAN( | |
7016d6eb DM |
1798 | is_LNBREAK_utf8_safe(s, strend), |
1799 | is_LNBREAK_latin1_safe(s, strend) | |
e1d1eefb | 1800 | ); |
6895a8aa | 1801 | break; |
e1d1eefb YO |
1802 | case VERTWS: |
1803 | REXEC_FBC_CSCAN( | |
7016d6eb DM |
1804 | is_VERTWS_utf8_safe(s, strend), |
1805 | is_VERTWS_latin1_safe(s, strend) | |
e1d1eefb | 1806 | ); |
6895a8aa | 1807 | break; |
e1d1eefb YO |
1808 | case NVERTWS: |
1809 | REXEC_FBC_CSCAN( | |
7016d6eb DM |
1810 | !is_VERTWS_utf8_safe(s, strend), |
1811 | !is_VERTWS_latin1_safe(s, strend) | |
e1d1eefb | 1812 | ); |
6895a8aa | 1813 | break; |
e1d1eefb YO |
1814 | case HORIZWS: |
1815 | REXEC_FBC_CSCAN( | |
7016d6eb DM |
1816 | is_HORIZWS_utf8_safe(s, strend), |
1817 | is_HORIZWS_latin1_safe(s, strend) | |
e1d1eefb | 1818 | ); |
6895a8aa | 1819 | break; |
e1d1eefb YO |
1820 | case NHORIZWS: |
1821 | REXEC_FBC_CSCAN( | |
7016d6eb DM |
1822 | !is_HORIZWS_utf8_safe(s, strend), |
1823 | !is_HORIZWS_latin1_safe(s, strend) | |
e1d1eefb | 1824 | ); |
6895a8aa | 1825 | break; |
0658cdde KW |
1826 | case POSIXA: |
1827 | /* Don't need to worry about utf8, as it can match only a single | |
1828 | * byte invariant character. The flag in this node type is the | |
1829 | * class number to pass to _generic_isCC() to build a mask for | |
1830 | * searching in PL_charclass[] */ | |
1831 | REXEC_FBC_CLASS_SCAN( _generic_isCC_A(*s, FLAGS(c))); | |
1832 | break; | |
1833 | case NPOSIXA: | |
1834 | REXEC_FBC_CSCAN( | |
1835 | !_generic_isCC_A(*s, FLAGS(c)), | |
1836 | !_generic_isCC_A(*s, FLAGS(c)) | |
1837 | ); | |
1838 | break; | |
1839 | ||
1de06328 YO |
1840 | case AHOCORASICKC: |
1841 | case AHOCORASICK: | |
07be1b83 | 1842 | { |
a0a388a1 | 1843 | DECL_TRIE_TYPE(c); |
07be1b83 YO |
1844 | /* what trie are we using right now */ |
1845 | reg_ac_data *aho | |
f8fc2ecf | 1846 | = (reg_ac_data*)progi->data->data[ ARG( c ) ]; |
3251b653 NC |
1847 | reg_trie_data *trie |
1848 | = (reg_trie_data*)progi->data->data[ aho->trie ]; | |
85fbaab2 | 1849 | HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]); |
07be1b83 YO |
1850 | |
1851 | const char *last_start = strend - trie->minlen; | |
6148ee25 | 1852 | #ifdef DEBUGGING |
07be1b83 | 1853 | const char *real_start = s; |
6148ee25 | 1854 | #endif |
07be1b83 | 1855 | STRLEN maxlen = trie->maxlen; |
be8e71aa YO |
1856 | SV *sv_points; |
1857 | U8 **points; /* map of where we were in the input string | |
786e8c11 | 1858 | when reading a given char. For ASCII this |
be8e71aa | 1859 | is unnecessary overhead as the relationship |
38a44b82 NC |
1860 | is always 1:1, but for Unicode, especially |
1861 | case folded Unicode this is not true. */ | |
f9e705e8 | 1862 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
786e8c11 YO |
1863 | U8 *bitmap=NULL; |
1864 | ||
07be1b83 YO |
1865 | |
1866 | GET_RE_DEBUG_FLAGS_DECL; | |
1867 | ||
be8e71aa YO |
1868 | /* We can't just allocate points here. We need to wrap it in |
1869 | * an SV so it gets freed properly if there is a croak while | |
1870 | * running the match */ | |
1871 | ENTER; | |
1872 | SAVETMPS; | |
1873 | sv_points=newSV(maxlen * sizeof(U8 *)); | |
1874 | SvCUR_set(sv_points, | |
1875 | maxlen * sizeof(U8 *)); | |
1876 | SvPOK_on(sv_points); | |
1877 | sv_2mortal(sv_points); | |
1878 | points=(U8**)SvPV_nolen(sv_points ); | |
1de06328 YO |
1879 | if ( trie_type != trie_utf8_fold |
1880 | && (trie->bitmap || OP(c)==AHOCORASICKC) ) | |
1881 | { | |
786e8c11 YO |
1882 | if (trie->bitmap) |
1883 | bitmap=(U8*)trie->bitmap; | |
1884 | else | |
1885 | bitmap=(U8*)ANYOF_BITMAP(c); | |
07be1b83 | 1886 | } |
786e8c11 YO |
1887 | /* this is the Aho-Corasick algorithm modified a touch |
1888 | to include special handling for long "unknown char" | |
1889 | sequences. The basic idea being that we use AC as long | |
1890 | as we are dealing with a possible matching char, when | |
1891 | we encounter an unknown char (and we have not encountered | |
1892 | an accepting state) we scan forward until we find a legal | |
1893 | starting char. | |
1894 | AC matching is basically that of trie matching, except | |
1895 | that when we encounter a failing transition, we fall back | |
1896 | to the current states "fail state", and try the current char | |
1897 | again, a process we repeat until we reach the root state, | |
1898 | state 1, or a legal transition. If we fail on the root state | |
1899 | then we can either terminate if we have reached an accepting | |
1900 | state previously, or restart the entire process from the beginning | |
1901 | if we have not. | |
1902 | ||
1903 | */ | |
07be1b83 YO |
1904 | while (s <= last_start) { |
1905 | const U32 uniflags = UTF8_ALLOW_DEFAULT; | |
1906 | U8 *uc = (U8*)s; | |
1907 | U16 charid = 0; | |
1908 | U32 base = 1; | |
1909 | U32 state = 1; | |
1910 | UV uvc = 0; | |
1911 | STRLEN len = 0; | |
1912 | STRLEN foldlen = 0; | |
1913 | U8 *uscan = (U8*)NULL; | |
1914 | U8 *leftmost = NULL; | |
786e8c11 YO |
1915 | #ifdef DEBUGGING |
1916 | U32 accepted_word= 0; | |
1917 | #endif | |
07be1b83 YO |
1918 | U32 pointpos = 0; |
1919 | ||
1920 | while ( state && uc <= (U8*)strend ) { | |
1921 | int failed=0; | |
786e8c11 YO |
1922 | U32 word = aho->states[ state ].wordnum; |
1923 | ||
1de06328 YO |
1924 | if( state==1 ) { |
1925 | if ( bitmap ) { | |
1926 | DEBUG_TRIE_EXECUTE_r( | |
1927 | if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1928 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1929 | (char *)uc, utf8_target ); |
1de06328 YO |
1930 | PerlIO_printf( Perl_debug_log, |
1931 | " Scanning for legal start char...\n"); | |
1932 | } | |
d085b490 YO |
1933 | ); |
1934 | if (utf8_target) { | |
1935 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1936 | uc += UTF8SKIP(uc); | |
1937 | } | |
1938 | } else { | |
1939 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1940 | uc++; | |
1941 | } | |
1942 | } | |
1de06328 | 1943 | s= (char *)uc; |
786e8c11 | 1944 | } |
786e8c11 YO |
1945 | if (uc >(U8*)last_start) break; |
1946 | } | |
1947 | ||
1948 | if ( word ) { | |
2e64971a | 1949 | U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ]; |
786e8c11 YO |
1950 | if (!leftmost || lpos < leftmost) { |
1951 | DEBUG_r(accepted_word=word); | |
07be1b83 | 1952 | leftmost= lpos; |
786e8c11 | 1953 | } |
07be1b83 | 1954 | if (base==0) break; |
786e8c11 | 1955 | |
07be1b83 YO |
1956 | } |
1957 | points[pointpos++ % maxlen]= uc; | |
7016d6eb DM |
1958 | if (foldlen || uc < (U8*)strend) { |
1959 | REXEC_TRIE_READ_CHAR(trie_type, trie, | |
1960 | widecharmap, uc, | |
55eed653 NC |
1961 | uscan, len, uvc, charid, foldlen, |
1962 | foldbuf, uniflags); | |
7016d6eb DM |
1963 | DEBUG_TRIE_EXECUTE_r({ |
1964 | dump_exec_pos( (char *)uc, c, strend, | |
1965 | real_start, s, utf8_target); | |
1966 | PerlIO_printf(Perl_debug_log, | |
1967 | " Charid:%3u CP:%4"UVxf" ", | |
1968 | charid, uvc); | |
1969 | }); | |
1970 | } | |
1971 | else { | |
1972 | len = 0; | |
1973 | charid = 0; | |
1974 | } | |
1975 | ||
07be1b83 YO |
1976 | |
1977 | do { | |
6148ee25 | 1978 | #ifdef DEBUGGING |
786e8c11 | 1979 | word = aho->states[ state ].wordnum; |
6148ee25 | 1980 | #endif |
07be1b83 YO |
1981 | base = aho->states[ state ].trans.base; |
1982 | ||
786e8c11 YO |
1983 | DEBUG_TRIE_EXECUTE_r({ |
1984 | if (failed) | |
1985 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1986 | s, utf8_target ); |
07be1b83 | 1987 | PerlIO_printf( Perl_debug_log, |
786e8c11 YO |
1988 | "%sState: %4"UVxf", word=%"UVxf, |
1989 | failed ? " Fail transition to " : "", | |
1990 | (UV)state, (UV)word); | |
1991 | }); | |
07be1b83 YO |
1992 | if ( base ) { |
1993 | U32 tmp; | |
6dd2be57 | 1994 | I32 offset; |
07be1b83 | 1995 | if (charid && |
6dd2be57 DM |
1996 | ( ((offset = base + charid |
1997 | - 1 - trie->uniquecharcount)) >= 0) | |
1998 | && ((U32)offset < trie->lasttrans) | |
1999 | && trie->trans[offset].check == state | |
2000 | && (tmp=trie->trans[offset].next)) | |
07be1b83 | 2001 | { |
786e8c11 YO |
2002 | DEBUG_TRIE_EXECUTE_r( |
2003 | PerlIO_printf( Perl_debug_log," - legal\n")); | |
07be1b83 YO |
2004 | state = tmp; |
2005 | break; | |
2006 | } | |
2007 | else { | |
786e8c11 YO |
2008 | DEBUG_TRIE_EXECUTE_r( |
2009 | PerlIO_printf( Perl_debug_log," - fail\n")); | |
2010 | failed = 1; | |
2011 | state = aho->fail[state]; | |
07be1b83 YO |
2012 | } |
2013 | } | |
2014 | else { | |
2015 | /* we must be accepting here */ | |
786e8c11 YO |
2016 | DEBUG_TRIE_EXECUTE_r( |
2017 | PerlIO_printf( Perl_debug_log," - accepting\n")); | |
2018 | failed = 1; | |
07be1b83 YO |
2019 | break; |
2020 | } | |
2021 | } while(state); | |
786e8c11 | 2022 | uc += len; |
07be1b83 YO |
2023 | if (failed) { |
2024 | if (leftmost) | |
2025 | break; | |
786e8c11 | 2026 | if (!state) state = 1; |
07be1b83 YO |
2027 | } |
2028 | } | |
2029 | if ( aho->states[ state ].wordnum ) { | |
2e64971a | 2030 | U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ]; |
786e8c11 YO |
2031 | if (!leftmost || lpos < leftmost) { |
2032 | DEBUG_r(accepted_word=aho->states[ state ].wordnum); | |
07be1b83 | 2033 | leftmost = lpos; |
786e8c11 | 2034 | } |
07be1b83 | 2035 | } |
07be1b83 YO |
2036 | if (leftmost) { |
2037 | s = (char*)leftmost; | |
786e8c11 YO |
2038 | DEBUG_TRIE_EXECUTE_r({ |
2039 | PerlIO_printf( | |
70685ca0 JH |
2040 | Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n", |
2041 | (UV)accepted_word, (IV)(s - real_start) | |
786e8c11 YO |
2042 | ); |
2043 | }); | |
24b23f37 | 2044 | if (!reginfo || regtry(reginfo, &s)) { |
be8e71aa YO |
2045 | FREETMPS; |
2046 | LEAVE; | |
07be1b83 | 2047 | goto got_it; |
be8e71aa | 2048 | } |
07be1b83 | 2049 | s = HOPc(s,1); |
786e8c11 YO |
2050 | DEBUG_TRIE_EXECUTE_r({ |
2051 | PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n"); | |
2052 | }); | |
07be1b83 | 2053 | } else { |
786e8c11 YO |
2054 | DEBUG_TRIE_EXECUTE_r( |
2055 | PerlIO_printf( Perl_debug_log,"No match.\n")); | |
07be1b83 YO |
2056 | break; |
2057 | } | |
2058 | } | |
be8e71aa YO |
2059 | FREETMPS; |
2060 | LEAVE; | |
07be1b83 YO |
2061 | } |
2062 | break; | |
b3c9acc1 | 2063 | default: |
3c3eec57 GS |
2064 | Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); |
2065 | break; | |
d6a28714 | 2066 | } |
6eb5f6b9 JH |
2067 | return 0; |
2068 | got_it: | |
2069 | return s; | |
2070 | } | |
2071 | ||
fae667d5 | 2072 | |
6eb5f6b9 JH |
2073 | /* |
2074 | - regexec_flags - match a regexp against a string | |
2075 | */ | |
2076 | I32 | |
288b8c02 | 2077 | Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend, |
6eb5f6b9 | 2078 | char *strbeg, I32 minend, SV *sv, void *data, U32 flags) |
8fd1a950 DM |
2079 | /* stringarg: the point in the string at which to begin matching */ |
2080 | /* strend: pointer to null at end of string */ | |
2081 | /* strbeg: real beginning of string */ | |
2082 | /* minend: end of match must be >= minend bytes after stringarg. */ | |
2083 | /* sv: SV being matched: only used for utf8 flag, pos() etc; string | |
2084 | * itself is accessed via the pointers above */ | |
2085 | /* data: May be used for some additional optimizations. | |
2086 | Currently its only used, with a U32 cast, for transmitting | |
2087 | the ganch offset when doing a /g match. This will change */ | |
2088 | /* nosave: For optimizations. */ | |
2089 | ||
6eb5f6b9 | 2090 | { |
97aff369 | 2091 | dVAR; |
288b8c02 | 2092 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
24b23f37 | 2093 | /*register*/ char *s; |
eb578fdb | 2094 | regnode *c; |
24b23f37 | 2095 | /*register*/ char *startpos = stringarg; |
6eb5f6b9 JH |
2096 | I32 minlen; /* must match at least this many chars */ |
2097 | I32 dontbother = 0; /* how many characters not to try at end */ | |
6eb5f6b9 JH |
2098 | I32 end_shift = 0; /* Same for the end. */ /* CC */ |
2099 | I32 scream_pos = -1; /* Internal iterator of scream. */ | |
ccac19ea | 2100 | char *scream_olds = NULL; |
f2ed9b32 | 2101 | const bool utf8_target = cBOOL(DO_UTF8(sv)); |
2757e526 | 2102 | I32 multiline; |
f8fc2ecf | 2103 | RXi_GET_DECL(prog,progi); |
3b0527fe | 2104 | regmatch_info reginfo; /* create some info to pass to regtry etc */ |
e9105d30 | 2105 | regexp_paren_pair *swap = NULL; |
a3621e74 YO |
2106 | GET_RE_DEBUG_FLAGS_DECL; |
2107 | ||
7918f24d | 2108 | PERL_ARGS_ASSERT_REGEXEC_FLAGS; |
9d4ba2ae | 2109 | PERL_UNUSED_ARG(data); |
6eb5f6b9 JH |
2110 | |
2111 | /* Be paranoid... */ | |
2112 | if (prog == NULL || startpos == NULL) { | |
2113 | Perl_croak(aTHX_ "NULL regexp parameter"); | |
2114 | return 0; | |
2115 | } | |
2116 | ||
bbe252da | 2117 | multiline = prog->extflags & RXf_PMf_MULTILINE; |
288b8c02 | 2118 | reginfo.prog = rx; /* Yes, sorry that this is confusing. */ |
2757e526 | 2119 | |
f2ed9b32 | 2120 | RX_MATCH_UTF8_set(rx, utf8_target); |
1de06328 | 2121 | DEBUG_EXECUTE_r( |
f2ed9b32 | 2122 | debug_start_match(rx, utf8_target, startpos, strend, |
1de06328 YO |
2123 | "Matching"); |
2124 | ); | |
bac06658 | 2125 | |
6eb5f6b9 | 2126 | minlen = prog->minlen; |
1de06328 YO |
2127 | |
2128 | if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) { | |
a3621e74 | 2129 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 JH |
2130 | "String too short [regexec_flags]...\n")); |
2131 | goto phooey; | |
1aa99e6b | 2132 | } |
6eb5f6b9 | 2133 | |
1de06328 | 2134 | |
6eb5f6b9 | 2135 | /* Check validity of program. */ |
f8fc2ecf | 2136 | if (UCHARAT(progi->program) != REG_MAGIC) { |
6eb5f6b9 JH |
2137 | Perl_croak(aTHX_ "corrupted regexp program"); |
2138 | } | |
2139 | ||
2140 | PL_reg_flags = 0; | |
ed301438 | 2141 | PL_reg_state.re_state_eval_setup_done = FALSE; |
6eb5f6b9 JH |
2142 | PL_reg_maxiter = 0; |
2143 | ||
3c8556c3 | 2144 | if (RX_UTF8(rx)) |
6eb5f6b9 JH |
2145 | PL_reg_flags |= RF_utf8; |
2146 | ||
2147 | /* Mark beginning of line for ^ and lookbehind. */ | |
3b0527fe | 2148 | reginfo.bol = startpos; /* XXX not used ??? */ |
6eb5f6b9 | 2149 | PL_bostr = strbeg; |
3b0527fe | 2150 | reginfo.sv = sv; |
6eb5f6b9 JH |
2151 | |
2152 | /* Mark end of line for $ (and such) */ | |
2153 | PL_regeol = strend; | |
2154 | ||
2155 | /* see how far we have to get to not match where we matched before */ | |
3b0527fe | 2156 | reginfo.till = startpos+minend; |
6eb5f6b9 | 2157 | |
6eb5f6b9 JH |
2158 | /* If there is a "must appear" string, look for it. */ |
2159 | s = startpos; | |
2160 | ||
bbe252da | 2161 | if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */ |
6eb5f6b9 | 2162 | MAGIC *mg; |
2c296965 | 2163 | if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */ |
58e23c8d | 2164 | reginfo.ganch = startpos + prog->gofs; |
2c296965 | 2165 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2166 | "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs)); |
2c296965 | 2167 | } else if (sv && SvTYPE(sv) >= SVt_PVMG |
6eb5f6b9 | 2168 | && SvMAGIC(sv) |
14befaf4 DM |
2169 | && (mg = mg_find(sv, PERL_MAGIC_regex_global)) |
2170 | && mg->mg_len >= 0) { | |
3b0527fe | 2171 | reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ |
2c296965 | 2172 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2173 | "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len)); |
2c296965 | 2174 | |
bbe252da | 2175 | if (prog->extflags & RXf_ANCH_GPOS) { |
3b0527fe | 2176 | if (s > reginfo.ganch) |
6eb5f6b9 | 2177 | goto phooey; |
58e23c8d | 2178 | s = reginfo.ganch - prog->gofs; |
2c296965 | 2179 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2180 | "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs)); |
c584a96e YO |
2181 | if (s < strbeg) |
2182 | goto phooey; | |
6eb5f6b9 JH |
2183 | } |
2184 | } | |
58e23c8d | 2185 | else if (data) { |
70685ca0 | 2186 | reginfo.ganch = strbeg + PTR2UV(data); |
2c296965 YO |
2187 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2188 | "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data))); | |
2189 | ||
2190 | } else { /* pos() not defined */ | |
3b0527fe | 2191 | reginfo.ganch = strbeg; |
2c296965 YO |
2192 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2193 | "GPOS: reginfo.ganch = strbeg\n")); | |
2194 | } | |
6eb5f6b9 | 2195 | } |
288b8c02 | 2196 | if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) { |
e9105d30 GG |
2197 | /* We have to be careful. If the previous successful match |
2198 | was from this regex we don't want a subsequent partially | |
2199 | successful match to clobber the old results. | |
2200 | So when we detect this possibility we add a swap buffer | |
2201 | to the re, and switch the buffer each match. If we fail | |
2202 | we switch it back, otherwise we leave it swapped. | |
2203 | */ | |
2204 | swap = prog->offs; | |
2205 | /* do we need a save destructor here for eval dies? */ | |
2206 | Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair); | |
495f47a5 DM |
2207 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
2208 | "rex=0x%"UVxf" saving offs: orig=0x%"UVxf" new=0x%"UVxf"\n", | |
2209 | PTR2UV(prog), | |
2210 | PTR2UV(swap), | |
2211 | PTR2UV(prog->offs) | |
2212 | )); | |
c74340f9 | 2213 | } |
a0714e2c | 2214 | if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { |
6eb5f6b9 JH |
2215 | re_scream_pos_data d; |
2216 | ||
2217 | d.scream_olds = &scream_olds; | |
2218 | d.scream_pos = &scream_pos; | |
288b8c02 | 2219 | s = re_intuit_start(rx, sv, s, strend, flags, &d); |
3fa9c3d7 | 2220 | if (!s) { |
a3621e74 | 2221 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n")); |
6eb5f6b9 | 2222 | goto phooey; /* not present */ |
3fa9c3d7 | 2223 | } |
6eb5f6b9 JH |
2224 | } |
2225 | ||
1de06328 | 2226 | |
6eb5f6b9 JH |
2227 | |
2228 | /* Simplest case: anchored match need be tried only once. */ | |
2229 | /* [unless only anchor is BOL and multiline is set] */ | |
bbe252da | 2230 | if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) { |
24b23f37 | 2231 | if (s == startpos && regtry(®info, &startpos)) |
6eb5f6b9 | 2232 | goto got_it; |
bbe252da YO |
2233 | else if (multiline || (prog->intflags & PREGf_IMPLICIT) |
2234 | || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */ | |
6eb5f6b9 JH |
2235 | { |
2236 | char *end; | |
2237 | ||
2238 | if (minlen) | |
2239 | dontbother = minlen - 1; | |
1aa99e6b | 2240 | end = HOP3c(strend, -dontbother, strbeg) - 1; |
6eb5f6b9 | 2241 | /* for multiline we only have to try after newlines */ |
33b8afdf | 2242 | if (prog->check_substr || prog->check_utf8) { |
92f3d482 YO |
2243 | /* because of the goto we can not easily reuse the macros for bifurcating the |
2244 | unicode/non-unicode match modes here like we do elsewhere - demerphq */ | |
2245 | if (utf8_target) { | |
2246 | if (s == startpos) | |
2247 | goto after_try_utf8; | |
2248 | while (1) { | |
2249 | if (regtry(®info, &s)) { | |
2250 | goto got_it; | |
2251 | } | |
2252 | after_try_utf8: | |
2253 | if (s > end) { | |
2254 | goto phooey; | |
2255 | } | |
2256 | if (prog->extflags & RXf_USE_INTUIT) { | |
2257 | s = re_intuit_start(rx, sv, s + UTF8SKIP(s), strend, flags, NULL); | |
2258 | if (!s) { | |
2259 | goto phooey; | |
2260 | } | |
2261 | } | |
2262 | else { | |
2263 | s += UTF8SKIP(s); | |
2264 | } | |
2265 | } | |
2266 | } /* end search for check string in unicode */ | |
2267 | else { | |
2268 | if (s == startpos) { | |
2269 | goto after_try_latin; | |
2270 | } | |
2271 | while (1) { | |
2272 | if (regtry(®info, &s)) { | |
2273 | goto got_it; | |
2274 | } | |
2275 | after_try_latin: | |
2276 | if (s > end) { | |
2277 | goto phooey; | |
2278 | } | |
2279 | if (prog->extflags & RXf_USE_INTUIT) { | |
2280 | s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL); | |
2281 | if (!s) { | |
2282 | goto phooey; | |
2283 | } | |
2284 | } | |
2285 | else { | |
2286 | s++; | |
2287 | } | |
2288 | } | |
2289 | } /* end search for check string in latin*/ | |
2290 | } /* end search for check string */ | |
2291 | else { /* search for newline */ | |
2292 | if (s > startpos) { | |
2293 | /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/ | |
6eb5f6b9 | 2294 | s--; |
92f3d482 | 2295 | } |
21eede78 YO |
2296 | /* We can use a more efficient search as newlines are the same in unicode as they are in latin */ |
2297 | while (s <= end) { /* note it could be possible to match at the end of the string */ | |
6eb5f6b9 | 2298 | if (*s++ == '\n') { /* don't need PL_utf8skip here */ |
24b23f37 | 2299 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2300 | goto got_it; |
2301 | } | |
92f3d482 YO |
2302 | } |
2303 | } /* end search for newline */ | |
2304 | } /* end anchored/multiline check string search */ | |
6eb5f6b9 | 2305 | goto phooey; |
bbe252da | 2306 | } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK)) |
f9f4320a | 2307 | { |
486ec47a | 2308 | /* the warning about reginfo.ganch being used without initialization |
bbe252da | 2309 | is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN |
f9f4320a | 2310 | and we only enter this block when the same bit is set. */ |
58e23c8d | 2311 | char *tmp_s = reginfo.ganch - prog->gofs; |
c584a96e YO |
2312 | |
2313 | if (tmp_s >= strbeg && regtry(®info, &tmp_s)) | |
6eb5f6b9 JH |
2314 | goto got_it; |
2315 | goto phooey; | |
2316 | } | |
2317 | ||
2318 | /* Messy cases: unanchored match. */ | |
bbe252da | 2319 | if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) { |
6eb5f6b9 | 2320 | /* we have /x+whatever/ */ |
f2ed9b32 | 2321 | /* it must be a one character string (XXXX Except UTF_PATTERN?) */ |
33b8afdf | 2322 | char ch; |
bf93d4cc GS |
2323 | #ifdef DEBUGGING |
2324 | int did_match = 0; | |
2325 | #endif | |
f2ed9b32 | 2326 | if (utf8_target) { |
7e0d5ad7 KW |
2327 | if (! prog->anchored_utf8) { |
2328 | to_utf8_substr(prog); | |
2329 | } | |
2330 | ch = SvPVX_const(prog->anchored_utf8)[0]; | |
4cadc6a9 | 2331 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2332 | if (*s == ch) { |
a3621e74 | 2333 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2334 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2335 | s += UTF8SKIP(s); |
2336 | while (s < strend && *s == ch) | |
2337 | s += UTF8SKIP(s); | |
2338 | } | |
4cadc6a9 | 2339 | ); |
7e0d5ad7 | 2340 | |
6eb5f6b9 JH |
2341 | } |
2342 | else { | |
7e0d5ad7 KW |
2343 | if (! prog->anchored_substr) { |
2344 | if (! to_byte_substr(prog)) { | |
2345 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
2346 | non_utf8_target_but_utf8_required)); | |
2347 | goto phooey; | |
2348 | } | |
2349 | } | |
2350 | ch = SvPVX_const(prog->anchored_substr)[0]; | |
4cadc6a9 | 2351 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2352 | if (*s == ch) { |
a3621e74 | 2353 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2354 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2355 | s++; |
2356 | while (s < strend && *s == ch) | |
2357 | s++; | |
2358 | } | |
4cadc6a9 | 2359 | ); |
6eb5f6b9 | 2360 | } |
a3621e74 | 2361 | DEBUG_EXECUTE_r(if (!did_match) |
bf93d4cc | 2362 | PerlIO_printf(Perl_debug_log, |
b7953727 JH |
2363 | "Did not find anchored character...\n") |
2364 | ); | |
6eb5f6b9 | 2365 | } |
a0714e2c SS |
2366 | else if (prog->anchored_substr != NULL |
2367 | || prog->anchored_utf8 != NULL | |
2368 | || ((prog->float_substr != NULL || prog->float_utf8 != NULL) | |
33b8afdf JH |
2369 | && prog->float_max_offset < strend - s)) { |
2370 | SV *must; | |
2371 | I32 back_max; | |
2372 | I32 back_min; | |
2373 | char *last; | |
6eb5f6b9 | 2374 | char *last1; /* Last position checked before */ |
bf93d4cc GS |
2375 | #ifdef DEBUGGING |
2376 | int did_match = 0; | |
2377 | #endif | |
33b8afdf | 2378 | if (prog->anchored_substr || prog->anchored_utf8) { |
7e0d5ad7 KW |
2379 | if (utf8_target) { |
2380 | if (! prog->anchored_utf8) { | |
2381 | to_utf8_substr(prog); | |
2382 | } | |
2383 | must = prog->anchored_utf8; | |
2384 | } | |
2385 | else { | |
2386 | if (! prog->anchored_substr) { | |
2387 | if (! to_byte_substr(prog)) { | |
2388 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
2389 | non_utf8_target_but_utf8_required)); | |
2390 | goto phooey; | |
2391 | } | |
2392 | } | |
2393 | must = prog->anchored_substr; | |
2394 | } | |
33b8afdf JH |
2395 | back_max = back_min = prog->anchored_offset; |
2396 | } else { | |
7e0d5ad7 KW |
2397 | if (utf8_target) { |
2398 | if (! prog->float_utf8) { | |
2399 | to_utf8_substr(prog); | |
2400 | } | |
2401 | must = prog->float_utf8; | |
2402 | } | |
2403 | else { | |
2404 | if (! prog->float_substr) { | |
2405 | if (! to_byte_substr(prog)) { | |
2406 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
2407 | non_utf8_target_but_utf8_required)); | |
2408 | goto phooey; | |
2409 | } | |
2410 | } | |
2411 | must = prog->float_substr; | |
2412 | } | |
33b8afdf JH |
2413 | back_max = prog->float_max_offset; |
2414 | back_min = prog->float_min_offset; | |
2415 | } | |
1de06328 | 2416 | |
1de06328 YO |
2417 | if (back_min<0) { |
2418 | last = strend; | |
2419 | } else { | |
2420 | last = HOP3c(strend, /* Cannot start after this */ | |
2421 | -(I32)(CHR_SVLEN(must) | |
2422 | - (SvTAIL(must) != 0) + back_min), strbeg); | |
2423 | } | |
6eb5f6b9 JH |
2424 | if (s > PL_bostr) |
2425 | last1 = HOPc(s, -1); | |
2426 | else | |
2427 | last1 = s - 1; /* bogus */ | |
2428 | ||
a0288114 | 2429 | /* XXXX check_substr already used to find "s", can optimize if |
6eb5f6b9 JH |
2430 | check_substr==must. */ |
2431 | scream_pos = -1; | |
2432 | dontbother = end_shift; | |
2433 | strend = HOPc(strend, -dontbother); | |
2434 | while ( (s <= last) && | |
c33e64f0 | 2435 | (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)), |
9041c2e3 | 2436 | (unsigned char*)strend, must, |
c33e64f0 | 2437 | multiline ? FBMrf_MULTILINE : 0)) ) { |
a3621e74 | 2438 | DEBUG_EXECUTE_r( did_match = 1 ); |
6eb5f6b9 JH |
2439 | if (HOPc(s, -back_max) > last1) { |
2440 | last1 = HOPc(s, -back_min); | |
2441 | s = HOPc(s, -back_max); | |
2442 | } | |
2443 | else { | |
52657f30 | 2444 | char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; |
6eb5f6b9 JH |
2445 | |
2446 | last1 = HOPc(s, -back_min); | |
52657f30 | 2447 | s = t; |
6eb5f6b9 | 2448 | } |
f2ed9b32 | 2449 | if (utf8_target) { |
6eb5f6b9 | 2450 | while (s <= last1) { |
24b23f37 | 2451 | if (regtry(®info, &s)) |
6eb5f6b9 | 2452 | goto got_it; |
7016d6eb DM |
2453 | if (s >= last1) { |
2454 | s++; /* to break out of outer loop */ | |
2455 | break; | |
2456 | } | |
2457 | s += UTF8SKIP(s); | |
6eb5f6b9 JH |
2458 | } |
2459 | } | |
2460 | else { | |
2461 | while (s <= last1) { | |
24b23f37 | 2462 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2463 | goto got_it; |
2464 | s++; | |
2465 | } | |
2466 | } | |
2467 | } | |
ab3bbdeb | 2468 | DEBUG_EXECUTE_r(if (!did_match) { |
f2ed9b32 | 2469 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
2470 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
2471 | PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n", | |
33b8afdf | 2472 | ((must == prog->anchored_substr || must == prog->anchored_utf8) |
bf93d4cc | 2473 | ? "anchored" : "floating"), |
ab3bbdeb YO |
2474 | quoted, RE_SV_TAIL(must)); |
2475 | }); | |
6eb5f6b9 JH |
2476 | goto phooey; |
2477 | } | |
f8fc2ecf | 2478 | else if ( (c = progi->regstclass) ) { |
f14c76ed | 2479 | if (minlen) { |
f8fc2ecf | 2480 | const OPCODE op = OP(progi->regstclass); |
66e933ab | 2481 | /* don't bother with what can't match */ |
786e8c11 | 2482 | if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE) |
f14c76ed RGS |
2483 | strend = HOPc(strend, -(minlen - 1)); |
2484 | } | |
a3621e74 | 2485 | DEBUG_EXECUTE_r({ |
be8e71aa | 2486 | SV * const prop = sv_newmortal(); |
32fc9b6a | 2487 | regprop(prog, prop, c); |
0df25f3d | 2488 | { |
f2ed9b32 | 2489 | RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb | 2490 | s,strend-s,60); |
0df25f3d | 2491 | PerlIO_printf(Perl_debug_log, |
1c8f8eb1 | 2492 | "Matching stclass %.*s against %s (%d bytes)\n", |
e4f74956 | 2493 | (int)SvCUR(prop), SvPVX_const(prop), |
ab3bbdeb | 2494 | quoted, (int)(strend - s)); |
0df25f3d | 2495 | } |
ffc61ed2 | 2496 | }); |
3b0527fe | 2497 | if (find_byclass(prog, c, s, strend, ®info)) |
6eb5f6b9 | 2498 | goto got_it; |
07be1b83 | 2499 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n")); |
d6a28714 JH |
2500 | } |
2501 | else { | |
2502 | dontbother = 0; | |
a0714e2c | 2503 | if (prog->float_substr != NULL || prog->float_utf8 != NULL) { |
33b8afdf | 2504 | /* Trim the end. */ |
6af40bd7 | 2505 | char *last= NULL; |
33b8afdf | 2506 | SV* float_real; |
c33e64f0 FC |
2507 | STRLEN len; |
2508 | const char *little; | |
33b8afdf | 2509 | |
7e0d5ad7 KW |
2510 | if (utf8_target) { |
2511 | if (! prog->float_utf8) { | |
2512 | to_utf8_substr(prog); | |
2513 | } | |
2514 | float_real = prog->float_utf8; | |
2515 | } | |
2516 | else { | |
2517 | if (! prog->float_substr) { | |
2518 | if (! to_byte_substr(prog)) { | |
2519 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
2520 | non_utf8_target_but_utf8_required)); | |
2521 | goto phooey; | |
2522 | } | |
2523 | } | |
2524 | float_real = prog->float_substr; | |
2525 | } | |
d6a28714 | 2526 | |
c33e64f0 FC |
2527 | little = SvPV_const(float_real, len); |
2528 | if (SvTAIL(float_real)) { | |
7f18ad16 KW |
2529 | /* This means that float_real contains an artificial \n on |
2530 | * the end due to the presence of something like this: | |
2531 | * /foo$/ where we can match both "foo" and "foo\n" at the | |
2532 | * end of the string. So we have to compare the end of the | |
2533 | * string first against the float_real without the \n and | |
2534 | * then against the full float_real with the string. We | |
2535 | * have to watch out for cases where the string might be | |
2536 | * smaller than the float_real or the float_real without | |
2537 | * the \n. */ | |
1a13b075 YO |
2538 | char *checkpos= strend - len; |
2539 | DEBUG_OPTIMISE_r( | |
2540 | PerlIO_printf(Perl_debug_log, | |
2541 | "%sChecking for float_real.%s\n", | |
2542 | PL_colors[4], PL_colors[5])); | |
2543 | if (checkpos + 1 < strbeg) { | |
7f18ad16 KW |
2544 | /* can't match, even if we remove the trailing \n |
2545 | * string is too short to match */ | |
1a13b075 YO |
2546 | DEBUG_EXECUTE_r( |
2547 | PerlIO_printf(Perl_debug_log, | |
2548 | "%sString shorter than required trailing substring, cannot match.%s\n", | |
2549 | PL_colors[4], PL_colors[5])); | |
2550 | goto phooey; | |
2551 | } else if (memEQ(checkpos + 1, little, len - 1)) { | |
7f18ad16 KW |
2552 | /* can match, the end of the string matches without the |
2553 | * "\n" */ | |
1a13b075 YO |
2554 | last = checkpos + 1; |
2555 | } else if (checkpos < strbeg) { | |
7f18ad16 KW |
2556 | /* cant match, string is too short when the "\n" is |
2557 | * included */ | |
1a13b075 YO |
2558 | DEBUG_EXECUTE_r( |
2559 | PerlIO_printf(Perl_debug_log, | |
2560 | "%sString does not contain required trailing substring, cannot match.%s\n", | |
2561 | PL_colors[4], PL_colors[5])); | |
2562 | goto phooey; | |
2563 | } else if (!multiline) { | |
7f18ad16 KW |
2564 | /* non multiline match, so compare with the "\n" at the |
2565 | * end of the string */ | |
1a13b075 YO |
2566 | if (memEQ(checkpos, little, len)) { |
2567 | last= checkpos; | |
2568 | } else { | |
2569 | DEBUG_EXECUTE_r( | |
2570 | PerlIO_printf(Perl_debug_log, | |
2571 | "%sString does not contain required trailing substring, cannot match.%s\n", | |
2572 | PL_colors[4], PL_colors[5])); | |
2573 | goto phooey; | |
2574 | } | |
2575 | } else { | |
7f18ad16 KW |
2576 | /* multiline match, so we have to search for a place |
2577 | * where the full string is located */ | |
d6a28714 | 2578 | goto find_last; |
1a13b075 | 2579 | } |
c33e64f0 | 2580 | } else { |
d6a28714 | 2581 | find_last: |
9041c2e3 | 2582 | if (len) |
d6a28714 | 2583 | last = rninstr(s, strend, little, little + len); |
b8c5462f | 2584 | else |
a0288114 | 2585 | last = strend; /* matching "$" */ |
b8c5462f | 2586 | } |
6af40bd7 | 2587 | if (!last) { |
7f18ad16 KW |
2588 | /* at one point this block contained a comment which was |
2589 | * probably incorrect, which said that this was a "should not | |
2590 | * happen" case. Even if it was true when it was written I am | |
2591 | * pretty sure it is not anymore, so I have removed the comment | |
2592 | * and replaced it with this one. Yves */ | |
6bda09f9 YO |
2593 | DEBUG_EXECUTE_r( |
2594 | PerlIO_printf(Perl_debug_log, | |
6af40bd7 YO |
2595 | "String does not contain required substring, cannot match.\n" |
2596 | )); | |
2597 | goto phooey; | |
bf93d4cc | 2598 | } |
d6a28714 JH |
2599 | dontbother = strend - last + prog->float_min_offset; |
2600 | } | |
2601 | if (minlen && (dontbother < minlen)) | |
2602 | dontbother = minlen - 1; | |
2603 | strend -= dontbother; /* this one's always in bytes! */ | |
2604 | /* We don't know much -- general case. */ | |
f2ed9b32 | 2605 | if (utf8_target) { |
d6a28714 | 2606 | for (;;) { |
24b23f37 | 2607 | if (regtry(®info, &s)) |
d6a28714 JH |
2608 | goto got_it; |
2609 | if (s >= strend) | |
2610 | break; | |
b8c5462f | 2611 | s += UTF8SKIP(s); |
d6a28714 JH |
2612 | }; |
2613 | } | |
2614 | else { | |
2615 | do { | |
24b23f37 | 2616 | if (regtry(®info, &s)) |
d6a28714 JH |
2617 | goto got_it; |
2618 | } while (s++ < strend); | |
2619 | } | |
2620 | } | |
2621 | ||
2622 | /* Failure. */ | |
2623 | goto phooey; | |
2624 | ||
2625 | got_it: | |
495f47a5 DM |
2626 | DEBUG_BUFFERS_r( |
2627 | if (swap) | |
2628 | PerlIO_printf(Perl_debug_log, | |
2629 | "rex=0x%"UVxf" freeing offs: 0x%"UVxf"\n", | |
2630 | PTR2UV(prog), | |
2631 | PTR2UV(swap) | |
2632 | ); | |
2633 | ); | |
e9105d30 | 2634 | Safefree(swap); |
288b8c02 | 2635 | RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted); |
d6a28714 | 2636 | |
ed301438 | 2637 | if (PL_reg_state.re_state_eval_setup_done) |
4f639d21 | 2638 | restore_pos(aTHX_ prog); |
5daac39c NC |
2639 | if (RXp_PAREN_NAMES(prog)) |
2640 | (void)hv_iterinit(RXp_PAREN_NAMES(prog)); | |
d6a28714 JH |
2641 | |
2642 | /* make sure $`, $&, $', and $digit will work later */ | |
2643 | if ( !(flags & REXEC_NOT_FIRST) ) { | |
d6a28714 | 2644 | if (flags & REXEC_COPY_STR) { |
f8c7b90f | 2645 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2646 | if ((SvIsCOW(sv) |
2647 | || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) { | |
2648 | if (DEBUG_C_TEST) { | |
2649 | PerlIO_printf(Perl_debug_log, | |
2650 | "Copy on write: regexp capture, type %d\n", | |
2651 | (int) SvTYPE(sv)); | |
2652 | } | |
77f8f7c1 | 2653 | RX_MATCH_COPY_FREE(rx); |
ed252734 | 2654 | prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv); |
d5263905 | 2655 | prog->subbeg = (char *)SvPVX_const(prog->saved_copy); |
ed252734 | 2656 | assert (SvPOKp(prog->saved_copy)); |
6502e081 DM |
2657 | prog->sublen = PL_regeol - strbeg; |
2658 | prog->suboffset = 0; | |
2659 | prog->subcoffset = 0; | |
ed252734 NC |
2660 | } else |
2661 | #endif | |
2662 | { | |
6502e081 DM |
2663 | I32 min = 0; |
2664 | I32 max = PL_regeol - strbeg; | |
2665 | I32 sublen; | |
2666 | ||
2667 | if ( (flags & REXEC_COPY_SKIP_POST) | |
2668 | && !(RX_EXTFLAGS(rx) & RXf_PMf_KEEPCOPY) /* //p */ | |
2669 | && !(PL_sawampersand & SAWAMPERSAND_RIGHT) | |
2670 | ) { /* don't copy $' part of string */ | |
3de645a8 | 2671 | U32 n = 0; |
6502e081 DM |
2672 | max = -1; |
2673 | /* calculate the right-most part of the string covered | |
2674 | * by a capture. Due to look-ahead, this may be to | |
2675 | * the right of $&, so we have to scan all captures */ | |
2676 | while (n <= prog->lastparen) { | |
2677 | if (prog->offs[n].end > max) | |
2678 | max = prog->offs[n].end; | |
2679 | n++; | |
2680 | } | |
2681 | if (max == -1) | |
2682 | max = (PL_sawampersand & SAWAMPERSAND_LEFT) | |
2683 | ? prog->offs[0].start | |
2684 | : 0; | |
2685 | assert(max >= 0 && max <= PL_regeol - strbeg); | |
2686 | } | |
2687 | ||
2688 | if ( (flags & REXEC_COPY_SKIP_PRE) | |
2689 | && !(RX_EXTFLAGS(rx) & RXf_PMf_KEEPCOPY) /* //p */ | |
2690 | && !(PL_sawampersand & SAWAMPERSAND_LEFT) | |
2691 | ) { /* don't copy $` part of string */ | |
3de645a8 | 2692 | U32 n = 0; |
6502e081 DM |
2693 | min = max; |
2694 | /* calculate the left-most part of the string covered | |
2695 | * by a capture. Due to look-behind, this may be to | |
2696 | * the left of $&, so we have to scan all captures */ | |
2697 | while (min && n <= prog->lastparen) { | |
2698 | if ( prog->offs[n].start != -1 | |
2699 | && prog->offs[n].start < min) | |
2700 | { | |
2701 | min = prog->offs[n].start; | |
2702 | } | |
2703 | n++; | |
2704 | } | |
2705 | if ((PL_sawampersand & SAWAMPERSAND_RIGHT) | |
2706 | && min > prog->offs[0].end | |
2707 | ) | |
2708 | min = prog->offs[0].end; | |
2709 | ||
2710 | } | |
2711 | ||
2712 | assert(min >= 0 && min <= max && min <= PL_regeol - strbeg); | |
2713 | sublen = max - min; | |
2714 | ||
2715 | if (RX_MATCH_COPIED(rx)) { | |
2716 | if (sublen > prog->sublen) | |
2717 | prog->subbeg = | |
2718 | (char*)saferealloc(prog->subbeg, sublen+1); | |
2719 | } | |
2720 | else | |
2721 | prog->subbeg = (char*)safemalloc(sublen+1); | |
2722 | Copy(strbeg + min, prog->subbeg, sublen, char); | |
2723 | prog->subbeg[sublen] = '\0'; | |
2724 | prog->suboffset = min; | |
2725 | prog->sublen = sublen; | |
77f8f7c1 | 2726 | RX_MATCH_COPIED_on(rx); |
6502e081 | 2727 | } |
6502e081 DM |
2728 | prog->subcoffset = prog->suboffset; |
2729 | if (prog->suboffset && utf8_target) { | |
2730 | /* Convert byte offset to chars. | |
2731 | * XXX ideally should only compute this if @-/@+ | |
2732 | * has been seen, a la PL_sawampersand ??? */ | |
2733 | ||
2734 | /* If there's a direct correspondence between the | |
2735 | * string which we're matching and the original SV, | |
2736 | * then we can use the utf8 len cache associated with | |
2737 | * the SV. In particular, it means that under //g, | |
2738 | * sv_pos_b2u() will use the previously cached | |
2739 | * position to speed up working out the new length of | |
2740 | * subcoffset, rather than counting from the start of | |
2741 | * the string each time. This stops | |
2742 | * $x = "\x{100}" x 1E6; 1 while $x =~ /(.)/g; | |
2743 | * from going quadratic */ | |
2744 | if (SvPOKp(sv) && SvPVX(sv) == strbeg) | |
2745 | sv_pos_b2u(sv, &(prog->subcoffset)); | |
2746 | else | |
2747 | prog->subcoffset = utf8_length((U8*)strbeg, | |
2748 | (U8*)(strbeg+prog->suboffset)); | |
2749 | } | |
d6a28714 JH |
2750 | } |
2751 | else { | |
6502e081 | 2752 | RX_MATCH_COPY_FREE(rx); |
d6a28714 | 2753 | prog->subbeg = strbeg; |
6502e081 DM |
2754 | prog->suboffset = 0; |
2755 | prog->subcoffset = 0; | |
d6a28714 JH |
2756 | prog->sublen = PL_regeol - strbeg; /* strend may have been modified */ |
2757 | } | |
2758 | } | |
9041c2e3 | 2759 | |
d6a28714 JH |
2760 | return 1; |
2761 | ||
2762 | phooey: | |
a3621e74 | 2763 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", |
e4584336 | 2764 | PL_colors[4], PL_colors[5])); |
ed301438 | 2765 | if (PL_reg_state.re_state_eval_setup_done) |
4f639d21 | 2766 | restore_pos(aTHX_ prog); |
e9105d30 | 2767 | if (swap) { |
c74340f9 | 2768 | /* we failed :-( roll it back */ |
495f47a5 DM |
2769 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
2770 | "rex=0x%"UVxf" rolling back offs: freeing=0x%"UVxf" restoring=0x%"UVxf"\n", | |
2771 | PTR2UV(prog), | |
2772 | PTR2UV(prog->offs), | |
2773 | PTR2UV(swap) | |
2774 | )); | |
e9105d30 GG |
2775 | Safefree(prog->offs); |
2776 | prog->offs = swap; | |
2777 | } | |
d6a28714 JH |
2778 | return 0; |
2779 | } | |
2780 | ||
6bda09f9 | 2781 | |
ec43f78b DM |
2782 | /* Set which rex is pointed to by PL_reg_state, handling ref counting. |
2783 | * Do inc before dec, in case old and new rex are the same */ | |
2784 | #define SET_reg_curpm(Re2) \ | |
2785 | if (PL_reg_state.re_state_eval_setup_done) { \ | |
2786 | (void)ReREFCNT_inc(Re2); \ | |
2787 | ReREFCNT_dec(PM_GETRE(PL_reg_curpm)); \ | |
2788 | PM_SETRE((PL_reg_curpm), (Re2)); \ | |
2789 | } | |
2790 | ||
2791 | ||
d6a28714 JH |
2792 | /* |
2793 | - regtry - try match at specific point | |
2794 | */ | |
2795 | STATIC I32 /* 0 failure, 1 success */ | |
f73aaa43 | 2796 | S_regtry(pTHX_ regmatch_info *reginfo, char **startposp) |
d6a28714 | 2797 | { |
97aff369 | 2798 | dVAR; |
d6a28714 | 2799 | CHECKPOINT lastcp; |
288b8c02 NC |
2800 | REGEXP *const rx = reginfo->prog; |
2801 | regexp *const prog = (struct regexp *)SvANY(rx); | |
f73aaa43 | 2802 | I32 result; |
f8fc2ecf | 2803 | RXi_GET_DECL(prog,progi); |
a3621e74 | 2804 | GET_RE_DEBUG_FLAGS_DECL; |
7918f24d NC |
2805 | |
2806 | PERL_ARGS_ASSERT_REGTRY; | |
2807 | ||
24b23f37 | 2808 | reginfo->cutpoint=NULL; |
d6a28714 | 2809 | |
ed301438 DM |
2810 | if ((prog->extflags & RXf_EVAL_SEEN) |
2811 | && !PL_reg_state.re_state_eval_setup_done) | |
2812 | { | |
d6a28714 JH |
2813 | MAGIC *mg; |
2814 | ||
ed301438 | 2815 | PL_reg_state.re_state_eval_setup_done = TRUE; |
3b0527fe | 2816 | if (reginfo->sv) { |
d6a28714 | 2817 | /* Make $_ available to executed code. */ |
3b0527fe | 2818 | if (reginfo->sv != DEFSV) { |
59f00321 | 2819 | SAVE_DEFSV; |
414bf5ae | 2820 | DEFSV_set(reginfo->sv); |
b8c5462f | 2821 | } |
d6a28714 | 2822 | |
3b0527fe DM |
2823 | if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) |
2824 | && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { | |
d6a28714 | 2825 | /* prepare for quick setting of pos */ |
d300d9fa | 2826 | #ifdef PERL_OLD_COPY_ON_WRITE |
51a9ea20 NC |
2827 | if (SvIsCOW(reginfo->sv)) |
2828 | sv_force_normal_flags(reginfo->sv, 0); | |
d300d9fa | 2829 | #endif |
3dab1dad | 2830 | mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global, |
d300d9fa | 2831 | &PL_vtbl_mglob, NULL, 0); |
d6a28714 | 2832 | mg->mg_len = -1; |
b8c5462f | 2833 | } |
d6a28714 JH |
2834 | PL_reg_magic = mg; |
2835 | PL_reg_oldpos = mg->mg_len; | |
4f639d21 | 2836 | SAVEDESTRUCTOR_X(restore_pos, prog); |
d6a28714 | 2837 | } |
09687e5a | 2838 | if (!PL_reg_curpm) { |
a02a5408 | 2839 | Newxz(PL_reg_curpm, 1, PMOP); |
09687e5a AB |
2840 | #ifdef USE_ITHREADS |
2841 | { | |
14a49a24 | 2842 | SV* const repointer = &PL_sv_undef; |
92313705 NC |
2843 | /* this regexp is also owned by the new PL_reg_curpm, which |
2844 | will try to free it. */ | |
d2ece331 | 2845 | av_push(PL_regex_padav, repointer); |
09687e5a AB |
2846 | PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav); |
2847 | PL_regex_pad = AvARRAY(PL_regex_padav); | |
2848 | } | |
2849 | #endif | |
2850 | } | |
ec43f78b | 2851 | SET_reg_curpm(rx); |
d6a28714 JH |
2852 | PL_reg_oldcurpm = PL_curpm; |
2853 | PL_curpm = PL_reg_curpm; | |
07bc277f | 2854 | if (RXp_MATCH_COPIED(prog)) { |
d6a28714 JH |
2855 | /* Here is a serious problem: we cannot rewrite subbeg, |
2856 | since it may be needed if this match fails. Thus | |
2857 | $` inside (?{}) could fail... */ | |
2858 | PL_reg_oldsaved = prog->subbeg; | |
2859 | PL_reg_oldsavedlen = prog->sublen; | |
6502e081 DM |
2860 | PL_reg_oldsavedoffset = prog->suboffset; |
2861 | PL_reg_oldsavedcoffset = prog->suboffset; | |
f8c7b90f | 2862 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2863 | PL_nrs = prog->saved_copy; |
2864 | #endif | |
07bc277f | 2865 | RXp_MATCH_COPIED_off(prog); |
d6a28714 JH |
2866 | } |
2867 | else | |
bd61b366 | 2868 | PL_reg_oldsaved = NULL; |
d6a28714 | 2869 | prog->subbeg = PL_bostr; |
6502e081 DM |
2870 | prog->suboffset = 0; |
2871 | prog->subcoffset = 0; | |
d6a28714 JH |
2872 | prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */ |
2873 | } | |
97ca13b7 | 2874 | #ifdef DEBUGGING |
f73aaa43 | 2875 | PL_reg_starttry = *startposp; |
97ca13b7 | 2876 | #endif |
f73aaa43 | 2877 | prog->offs[0].start = *startposp - PL_bostr; |
d6a28714 | 2878 | prog->lastparen = 0; |
03994de8 | 2879 | prog->lastcloseparen = 0; |
d6a28714 | 2880 | PL_regsize = 0; |
d6a28714 JH |
2881 | |
2882 | /* XXXX What this code is doing here?!!! There should be no need | |
b93070ed | 2883 | to do this again and again, prog->lastparen should take care of |
3dd2943c | 2884 | this! --ilya*/ |
dafc8851 JH |
2885 | |
2886 | /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code. | |
2887 | * Actually, the code in regcppop() (which Ilya may be meaning by | |
b93070ed | 2888 | * prog->lastparen), is not needed at all by the test suite |
225593e1 DM |
2889 | * (op/regexp, op/pat, op/split), but that code is needed otherwise |
2890 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ | |
2891 | * Meanwhile, this code *is* needed for the | |
daf18116 JH |
2892 | * above-mentioned test suite tests to succeed. The common theme |
2893 | * on those tests seems to be returning null fields from matches. | |
225593e1 | 2894 | * --jhi updated by dapm */ |
dafc8851 | 2895 | #if 1 |
d6a28714 | 2896 | if (prog->nparens) { |
b93070ed | 2897 | regexp_paren_pair *pp = prog->offs; |
eb578fdb | 2898 | I32 i; |
b93070ed | 2899 | for (i = prog->nparens; i > (I32)prog->lastparen; i--) { |
f0ab9afb NC |
2900 | ++pp; |
2901 | pp->start = -1; | |
2902 | pp->end = -1; | |
d6a28714 JH |
2903 | } |
2904 | } | |
dafc8851 | 2905 | #endif |
02db2b7b | 2906 | REGCP_SET(lastcp); |
f73aaa43 DM |
2907 | result = regmatch(reginfo, *startposp, progi->program + 1); |
2908 | if (result != -1) { | |
2909 | prog->offs[0].end = result; | |
d6a28714 JH |
2910 | return 1; |
2911 | } | |
24b23f37 | 2912 | if (reginfo->cutpoint) |
f73aaa43 | 2913 | *startposp= reginfo->cutpoint; |
02db2b7b | 2914 | REGCP_UNWIND(lastcp); |
d6a28714 JH |
2915 | return 0; |
2916 | } | |
2917 | ||
02db2b7b | 2918 | |
8ba1375e MJD |
2919 | #define sayYES goto yes |
2920 | #define sayNO goto no | |
262b90c4 | 2921 | #define sayNO_SILENT goto no_silent |
8ba1375e | 2922 | |
f9f4320a YO |
2923 | /* we dont use STMT_START/END here because it leads to |
2924 | "unreachable code" warnings, which are bogus, but distracting. */ | |
2925 | #define CACHEsayNO \ | |
c476f425 DM |
2926 | if (ST.cache_mask) \ |
2927 | PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \ | |
f9f4320a | 2928 | sayNO |
3298f257 | 2929 | |
a3621e74 | 2930 | /* this is used to determine how far from the left messages like |
265c4333 YO |
2931 | 'failed...' are printed. It should be set such that messages |
2932 | are inline with the regop output that created them. | |
a3621e74 | 2933 | */ |
265c4333 | 2934 | #define REPORT_CODE_OFF 32 |
a3621e74 YO |
2935 | |
2936 | ||
40a82448 DM |
2937 | #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */ |
2938 | #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */ | |
79a2a0e8 KW |
2939 | #define CHRTEST_NOT_A_CP_1 -999 |
2940 | #define CHRTEST_NOT_A_CP_2 -998 | |
9e137952 | 2941 | |
86545054 DM |
2942 | #define SLAB_FIRST(s) (&(s)->states[0]) |
2943 | #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) | |
2944 | ||
5d9a96ca DM |
2945 | /* grab a new slab and return the first slot in it */ |
2946 | ||
2947 | STATIC regmatch_state * | |
2948 | S_push_slab(pTHX) | |
2949 | { | |
a35a87e7 | 2950 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
2951 | dMY_CXT; |
2952 | #endif | |
5d9a96ca DM |
2953 | regmatch_slab *s = PL_regmatch_slab->next; |
2954 | if (!s) { | |
2955 | Newx(s, 1, regmatch_slab); | |
2956 | s->prev = PL_regmatch_slab; | |
2957 | s->next = NULL; | |
2958 | PL_regmatch_slab->next = s; | |
2959 | } | |
2960 | PL_regmatch_slab = s; | |
86545054 | 2961 | return SLAB_FIRST(s); |
5d9a96ca | 2962 | } |
5b47454d | 2963 | |
95b24440 | 2964 | |
40a82448 DM |
2965 | /* push a new state then goto it */ |
2966 | ||
4d5016e5 DM |
2967 | #define PUSH_STATE_GOTO(state, node, input) \ |
2968 | pushinput = input; \ | |
40a82448 DM |
2969 | scan = node; \ |
2970 | st->resume_state = state; \ | |
2971 | goto push_state; | |
2972 | ||
2973 | /* push a new state with success backtracking, then goto it */ | |
2974 | ||
4d5016e5 DM |
2975 | #define PUSH_YES_STATE_GOTO(state, node, input) \ |
2976 | pushinput = input; \ | |
40a82448 DM |
2977 | scan = node; \ |
2978 | st->resume_state = state; \ | |
2979 | goto push_yes_state; | |
2980 | ||
aa283a38 | 2981 | |
aa283a38 | 2982 | |
4d5016e5 | 2983 | |
d6a28714 | 2984 | /* |
95b24440 | 2985 | |
bf1f174e DM |
2986 | regmatch() - main matching routine |
2987 | ||
2988 | This is basically one big switch statement in a loop. We execute an op, | |
2989 | set 'next' to point the next op, and continue. If we come to a point which | |
2990 | we may need to backtrack to on failure such as (A|B|C), we push a | |
2991 | backtrack state onto the backtrack stack. On failure, we pop the top | |
2992 | state, and re-enter the loop at the state indicated. If there are no more | |
2993 | states to pop, we return failure. | |
2994 | ||
2995 | Sometimes we also need to backtrack on success; for example /A+/, where | |
2996 | after successfully matching one A, we need to go back and try to | |
2997 | match another one; similarly for lookahead assertions: if the assertion | |
2998 | completes successfully, we backtrack to the state just before the assertion | |
2999 | and then carry on. In these cases, the pushed state is marked as | |
3000 | 'backtrack on success too'. This marking is in fact done by a chain of | |
3001 | pointers, each pointing to the previous 'yes' state. On success, we pop to | |
3002 | the nearest yes state, discarding any intermediate failure-only states. | |
3003 | Sometimes a yes state is pushed just to force some cleanup code to be | |
3004 | called at the end of a successful match or submatch; e.g. (??{$re}) uses | |
3005 | it to free the inner regex. | |
3006 | ||
3007 | Note that failure backtracking rewinds the cursor position, while | |
3008 | success backtracking leaves it alone. | |
3009 | ||
3010 | A pattern is complete when the END op is executed, while a subpattern | |
3011 | such as (?=foo) is complete when the SUCCESS op is executed. Both of these | |
3012 | ops trigger the "pop to last yes state if any, otherwise return true" | |
3013 | behaviour. | |
3014 | ||
3015 | A common convention in this function is to use A and B to refer to the two | |
3016 | subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is | |
3017 | the subpattern to be matched possibly multiple times, while B is the entire | |
3018 | rest of the pattern. Variable and state names reflect this convention. | |
3019 | ||
3020 | The states in the main switch are the union of ops and failure/success of | |
3021 | substates associated with with that op. For example, IFMATCH is the op | |
3022 | that does lookahead assertions /(?=A)B/ and so the IFMATCH state means | |
3023 | 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just | |
3024 | successfully matched A and IFMATCH_A_fail is a state saying that we have | |
3025 | just failed to match A. Resume states always come in pairs. The backtrack | |
3026 | state we push is marked as 'IFMATCH_A', but when that is popped, we resume | |
3027 | at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking | |
3028 | on success or failure. | |
3029 | ||
3030 | The struct that holds a backtracking state is actually a big union, with | |
3031 | one variant for each major type of op. The variable st points to the | |
3032 | top-most backtrack struct. To make the code clearer, within each | |
3033 | block of code we #define ST to alias the relevant union. | |
3034 | ||
3035 | Here's a concrete example of a (vastly oversimplified) IFMATCH | |
3036 | implementation: | |
3037 | ||
3038 | switch (state) { | |
3039 | .... | |
3040 | ||
3041 | #define ST st->u.ifmatch | |
3042 | ||
3043 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B | |
3044 | ST.foo = ...; // some state we wish to save | |
95b24440 | 3045 | ... |
bf1f174e DM |
3046 | // push a yes backtrack state with a resume value of |
3047 | // IFMATCH_A/IFMATCH_A_fail, then continue execution at the | |
3048 | // first node of A: | |
4d5016e5 | 3049 | PUSH_YES_STATE_GOTO(IFMATCH_A, A, newinput); |
bf1f174e DM |
3050 | // NOTREACHED |
3051 | ||
3052 | case IFMATCH_A: // we have successfully executed A; now continue with B | |
3053 | next = B; | |
3054 | bar = ST.foo; // do something with the preserved value | |
3055 | break; | |
3056 | ||
3057 | case IFMATCH_A_fail: // A failed, so the assertion failed | |
3058 | ...; // do some housekeeping, then ... | |
3059 | sayNO; // propagate the failure | |
3060 | ||
3061 | #undef ST | |
95b24440 | 3062 | |
bf1f174e DM |
3063 | ... |
3064 | } | |
95b24440 | 3065 | |
bf1f174e DM |
3066 | For any old-timers reading this who are familiar with the old recursive |
3067 | approach, the code above is equivalent to: | |
95b24440 | 3068 | |
bf1f174e DM |
3069 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B |
3070 | { | |
3071 | int foo = ... | |
95b24440 | 3072 | ... |
bf1f174e DM |
3073 | if (regmatch(A)) { |
3074 | next = B; | |
3075 | bar = foo; | |
3076 | break; | |
95b24440 | 3077 | } |
bf1f174e DM |
3078 | ...; // do some housekeeping, then ... |
3079 | sayNO; // propagate the failure | |
95b24440 | 3080 | } |
bf1f174e DM |
3081 | |
3082 | The topmost backtrack state, pointed to by st, is usually free. If you | |
3083 | want to claim it, populate any ST.foo fields in it with values you wish to | |
3084 | save, then do one of | |
3085 | ||
4d5016e5 DM |
3086 | PUSH_STATE_GOTO(resume_state, node, newinput); |
3087 | PUSH_YES_STATE_GOTO(resume_state, node, newinput); | |
bf1f174e DM |
3088 | |
3089 | which sets that backtrack state's resume value to 'resume_state', pushes a | |
3090 | new free entry to the top of the backtrack stack, then goes to 'node'. | |
3091 | On backtracking, the free slot is popped, and the saved state becomes the | |
3092 | new free state. An ST.foo field in this new top state can be temporarily | |
3093 | accessed to retrieve values, but once the main loop is re-entered, it | |
3094 | becomes available for reuse. | |
3095 | ||
3096 | Note that the depth of the backtrack stack constantly increases during the | |
3097 | left-to-right execution of the pattern, rather than going up and down with | |
3098 | the pattern nesting. For example the stack is at its maximum at Z at the | |
3099 | end of the pattern, rather than at X in the following: | |
3100 | ||
3101 | /(((X)+)+)+....(Y)+....Z/ | |
3102 | ||
3103 | The only exceptions to this are lookahead/behind assertions and the cut, | |
3104 | (?>A), which pop all the backtrack states associated with A before | |
3105 | continuing. | |
3106 | ||
486ec47a | 3107 | Backtrack state structs are allocated in slabs of about 4K in size. |
bf1f174e DM |
3108 | PL_regmatch_state and st always point to the currently active state, |
3109 | and PL_regmatch_slab points to the slab currently containing | |
3110 | PL_regmatch_state. The first time regmatch() is called, the first slab is | |
3111 | allocated, and is never freed until interpreter destruction. When the slab | |
3112 | is full, a new one is allocated and chained to the end. At exit from | |
3113 | regmatch(), slabs allocated since entry are freed. | |
3114 | ||
3115 | */ | |
95b24440 | 3116 | |
40a82448 | 3117 | |
5bc10b2c | 3118 | #define DEBUG_STATE_pp(pp) \ |
265c4333 | 3119 | DEBUG_STATE_r({ \ |
f2ed9b32 | 3120 | DUMP_EXEC_POS(locinput, scan, utf8_target); \ |
5bc10b2c | 3121 | PerlIO_printf(Perl_debug_log, \ |
5d458dd8 | 3122 | " %*s"pp" %s%s%s%s%s\n", \ |
5bc10b2c | 3123 | depth*2, "", \ |
13d6edb4 | 3124 | PL_reg_name[st->resume_state], \ |
5d458dd8 YO |
3125 | ((st==yes_state||st==mark_state) ? "[" : ""), \ |
3126 | ((st==yes_state) ? "Y" : ""), \ | |
3127 | ((st==mark_state) ? "M" : ""), \ | |
3128 | ((st==yes_state||st==mark_state) ? "]" : "") \ | |
3129 | ); \ | |
265c4333 | 3130 | }); |
5bc10b2c | 3131 | |
40a82448 | 3132 | |
3dab1dad | 3133 | #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1) |
95b24440 | 3134 | |
3df15adc | 3135 | #ifdef DEBUGGING |
5bc10b2c | 3136 | |
ab3bbdeb | 3137 | STATIC void |
f2ed9b32 | 3138 | S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target, |
ab3bbdeb YO |
3139 | const char *start, const char *end, const char *blurb) |
3140 | { | |
efd26800 | 3141 | const bool utf8_pat = RX_UTF8(prog) ? 1 : 0; |
7918f24d NC |
3142 | |
3143 | PERL_ARGS_ASSERT_DEBUG_START_MATCH; | |
3144 | ||
ab3bbdeb YO |
3145 | if (!PL_colorset) |
3146 | reginitcolors(); | |
3147 | { | |
3148 | RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0), | |
d2c6dc5e | 3149 | RX_PRECOMP_const(prog), RX_PRELEN(prog), 60); |
ab3bbdeb | 3150 | |
f2ed9b32 | 3151 | RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb YO |
3152 | start, end - start, 60); |
3153 | ||
3154 | PerlIO_printf(Perl_debug_log, | |
3155 | "%s%s REx%s %s against %s\n", | |
3156 | PL_colors[4], blurb, PL_colors[5], s0, s1); | |
3157 | ||
f2ed9b32 | 3158 | if (utf8_target||utf8_pat) |
1de06328 YO |
3159 | PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n", |
3160 | utf8_pat ? "pattern" : "", | |
f2ed9b32 KW |
3161 | utf8_pat && utf8_target ? " and " : "", |
3162 | utf8_target ? "string" : "" | |
ab3bbdeb YO |
3163 | ); |
3164 | } | |
3165 | } | |
3df15adc YO |
3166 | |
3167 | STATIC void | |
786e8c11 YO |
3168 | S_dump_exec_pos(pTHX_ const char *locinput, |
3169 | const regnode *scan, | |
3170 | const char *loc_regeol, | |
3171 | const char *loc_bostr, | |
3172 | const char *loc_reg_starttry, | |
f2ed9b32 | 3173 | const bool utf8_target) |
07be1b83 | 3174 | { |
786e8c11 | 3175 | const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4]; |
07be1b83 | 3176 | const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ |
786e8c11 | 3177 | int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput); |
07be1b83 YO |
3178 | /* The part of the string before starttry has one color |
3179 | (pref0_len chars), between starttry and current | |
3180 | position another one (pref_len - pref0_len chars), | |
3181 | after the current position the third one. | |
3182 | We assume that pref0_len <= pref_len, otherwise we | |
3183 | decrease pref0_len. */ | |
786e8c11 YO |
3184 | int pref_len = (locinput - loc_bostr) > (5 + taill) - l |
3185 | ? (5 + taill) - l : locinput - loc_bostr; | |
07be1b83 YO |
3186 | int pref0_len; |
3187 | ||
7918f24d NC |
3188 | PERL_ARGS_ASSERT_DUMP_EXEC_POS; |
3189 | ||
f2ed9b32 | 3190 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) |
07be1b83 | 3191 | pref_len++; |
786e8c11 YO |
3192 | pref0_len = pref_len - (locinput - loc_reg_starttry); |
3193 | if (l + pref_len < (5 + taill) && l < loc_regeol - locinput) | |
3194 | l = ( loc_regeol - locinput > (5 + taill) - pref_len | |
3195 | ? (5 + taill) - pref_len : loc_regeol - locinput); | |
f2ed9b32 | 3196 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) |
07be1b83 YO |
3197 | l--; |
3198 | if (pref0_len < 0) | |
3199 | pref0_len = 0; | |
3200 | if (pref0_len > pref_len) | |
3201 | pref0_len = pref_len; | |
3202 | { | |
f2ed9b32 | 3203 | const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0; |
0df25f3d | 3204 | |
ab3bbdeb | 3205 | RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0), |
1de06328 | 3206 | (locinput - pref_len),pref0_len, 60, 4, 5); |
0df25f3d | 3207 | |
ab3bbdeb | 3208 | RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1), |
3df15adc | 3209 | (locinput - pref_len + pref0_len), |
1de06328 | 3210 | pref_len - pref0_len, 60, 2, 3); |
0df25f3d | 3211 | |
ab3bbdeb | 3212 | RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2), |
1de06328 | 3213 | locinput, loc_regeol - locinput, 10, 0, 1); |
0df25f3d | 3214 | |
1de06328 | 3215 | const STRLEN tlen=len0+len1+len2; |
3df15adc | 3216 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb | 3217 | "%4"IVdf" <%.*s%.*s%s%.*s>%*s|", |
786e8c11 | 3218 | (IV)(locinput - loc_bostr), |
07be1b83 | 3219 | len0, s0, |
07be1b83 | 3220 | len1, s1, |
07be1b83 | 3221 | (docolor ? "" : "> <"), |
07be1b83 | 3222 | len2, s2, |
f9f4320a | 3223 | (int)(tlen > 19 ? 0 : 19 - tlen), |
07be1b83 YO |
3224 | ""); |
3225 | } | |
3226 | } | |
3df15adc | 3227 | |
07be1b83 YO |
3228 | #endif |
3229 | ||
0a4db386 YO |
3230 | /* reg_check_named_buff_matched() |
3231 | * Checks to see if a named buffer has matched. The data array of | |
3232 | * buffer numbers corresponding to the buffer is expected to reside | |
3233 | * in the regexp->data->data array in the slot stored in the ARG() of | |
3234 | * node involved. Note that this routine doesn't actually care about the | |
3235 | * name, that information is not preserved from compilation to execution. | |
3236 | * Returns the index of the leftmost defined buffer with the given name | |
3237 | * or 0 if non of the buffers matched. | |
3238 | */ | |
3239 | STATIC I32 | |
7918f24d NC |
3240 | S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) |
3241 | { | |
0a4db386 | 3242 | I32 n; |
f8fc2ecf | 3243 | RXi_GET_DECL(rex,rexi); |
ad64d0ec | 3244 | SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
0a4db386 | 3245 | I32 *nums=(I32*)SvPVX(sv_dat); |
7918f24d NC |
3246 | |
3247 | PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED; | |
3248 | ||
0a4db386 | 3249 | for ( n=0; n<SvIVX(sv_dat); n++ ) { |
b93070ed DM |
3250 | if ((I32)rex->lastparen >= nums[n] && |
3251 | rex->offs[nums[n]].end != -1) | |
0a4db386 YO |
3252 | { |
3253 | return nums[n]; | |
3254 | } | |
3255 | } | |
3256 | return 0; | |
3257 | } | |
3258 | ||
2f554ef7 DM |
3259 | |
3260 | /* free all slabs above current one - called during LEAVE_SCOPE */ | |
3261 | ||
3262 | STATIC void | |
3263 | S_clear_backtrack_stack(pTHX_ void *p) | |
3264 | { | |
3265 | regmatch_slab *s = PL_regmatch_slab->next; | |
3266 | PERL_UNUSED_ARG(p); | |
3267 | ||
3268 | if (!s) | |
3269 | return; | |
3270 | PL_regmatch_slab->next = NULL; | |
3271 | while (s) { | |
3272 | regmatch_slab * const osl = s; | |
3273 | s = s->next; | |
3274 | Safefree(osl); | |
3275 | } | |
3276 | } | |
c74f6de9 | 3277 | static bool |
79a2a0e8 | 3278 | S_setup_EXACTISH_ST_c1_c2(pTHX_ const regnode * const text_node, int *c1p, U8* c1_utf8, int *c2p, U8* c2_utf8) |
c74f6de9 | 3279 | { |
79a2a0e8 KW |
3280 | /* This function determines if there are one or two characters that match |
3281 | * the first character of the passed-in EXACTish node <text_node>, and if | |
3282 | * so, returns them in the passed-in pointers. | |
c74f6de9 | 3283 | * |
79a2a0e8 KW |
3284 | * If it determines that no possible character in the target string can |
3285 | * match, it returns FALSE; otherwise TRUE. (The FALSE situation occurs if | |
3286 | * the first character in <text_node> requires UTF-8 to represent, and the | |
3287 | * target string isn't in UTF-8.) | |
c74f6de9 | 3288 | * |
79a2a0e8 KW |
3289 | * If there are more than two characters that could match the beginning of |
3290 | * <text_node>, or if more context is required to determine a match or not, | |
3291 | * it sets both *<c1p> and *<c2p> to CHRTEST_VOID. | |
3292 | * | |
3293 | * The motiviation behind this function is to allow the caller to set up | |
3294 | * tight loops for matching. If <text_node> is of type EXACT, there is | |
3295 | * only one possible character that can match its first character, and so | |
3296 | * the situation is quite simple. But things get much more complicated if | |
3297 | * folding is involved. It may be that the first character of an EXACTFish | |
3298 | * node doesn't participate in any possible fold, e.g., punctuation, so it | |
3299 | * can be matched only by itself. The vast majority of characters that are | |
3300 | * in folds match just two things, their lower and upper-case equivalents. | |
3301 | * But not all are like that; some have multiple possible matches, or match | |
3302 | * sequences of more than one character. This function sorts all that out. | |
3303 | * | |
3304 | * Consider the patterns A*B or A*?B where A and B are arbitrary. In a | |
3305 | * loop of trying to match A*, we know we can't exit where the thing | |
3306 | * following it isn't a B. And something can't be a B unless it is the | |
3307 | * beginning of B. By putting a quick test for that beginning in a tight | |
3308 | * loop, we can rule out things that can't possibly be B without having to | |
3309 | * break out of the loop, thus avoiding work. Similarly, if A is a single | |
3310 | * character, we can make a tight loop matching A*, using the outputs of | |
3311 | * this function. | |
3312 | * | |
3313 | * If the target string to match isn't in UTF-8, and there aren't | |
3314 | * complications which require CHRTEST_VOID, *<c1p> and *<c2p> are set to | |
3315 | * the one or two possible octets (which are characters in this situation) | |
3316 | * that can match. In all cases, if there is only one character that can | |
3317 | * match, *<c1p> and *<c2p> will be identical. | |
3318 | * | |
3319 | * If the target string is in UTF-8, the buffers pointed to by <c1_utf8> | |
3320 | * and <c2_utf8> will contain the one or two UTF-8 sequences of bytes that | |
3321 | * can match the beginning of <text_node>. They should be declared with at | |
3322 | * least length UTF8_MAXBYTES+1. (If the target string isn't in UTF-8, it is | |
3323 | * undefined what these contain.) If one or both of the buffers are | |
3324 | * invariant under UTF-8, *<c1p>, and *<c2p> will also be set to the | |
3325 | * corresponding invariant. If variant, the corresponding *<c1p> and/or | |
3326 | * *<c2p> will be set to a negative number(s) that shouldn't match any code | |
3327 | * point (unless inappropriately coerced to unsigned). *<c1p> will equal | |
3328 | * *<c2p> if and only if <c1_utf8> and <c2_utf8> are the same. */ | |
c74f6de9 KW |
3329 | |
3330 | const bool utf8_target = PL_reg_match_utf8; | |
79a2a0e8 KW |
3331 | |
3332 | UV c1, c2; | |
3333 | bool use_chrtest_void = FALSE; | |
3334 | ||
3335 | /* Used when we have both utf8 input and utf8 output, to avoid converting | |
3336 | * to/from code points */ | |
3337 | bool utf8_has_been_setup = FALSE; | |
3338 | ||
c74f6de9 KW |
3339 | dVAR; |
3340 | ||
b4291290 | 3341 | U8 *pat = (U8*)STRING(text_node); |
c74f6de9 | 3342 | |
79a2a0e8 KW |
3343 | if (OP(text_node) == EXACT) { |
3344 | ||
3345 | /* In an exact node, only one thing can be matched, that first | |
3346 | * character. If both the pat and the target are UTF-8, we can just | |
3347 | * copy the input to the output, avoiding finding the code point of | |
3348 | * that character */ | |
3349 | if (! UTF_PATTERN) { | |
3350 | c2 = c1 = *pat; | |
3351 | } | |
3352 | else if (utf8_target) { | |
3353 | Copy(pat, c1_utf8, UTF8SKIP(pat), U8); | |
3354 | Copy(pat, c2_utf8, UTF8SKIP(pat), U8); | |
3355 | utf8_has_been_setup = TRUE; | |
3356 | } | |
3357 | else { | |
3358 | c2 = c1 = valid_utf8_to_uvchr(pat, NULL); | |
c74f6de9 | 3359 | } |
79a2a0e8 KW |
3360 | } |
3361 | else /* an EXACTFish node */ | |
3362 | if ((UTF_PATTERN | |
3363 | && is_MULTI_CHAR_FOLD_utf8_safe(pat, | |
3364 | pat + STR_LEN(text_node))) | |
3365 | || (! UTF_PATTERN | |
3366 | && is_MULTI_CHAR_FOLD_latin1_safe(pat, | |
3367 | pat + STR_LEN(text_node)))) | |
3368 | { | |
3369 | /* Multi-character folds require more context to sort out. Also | |
3370 | * PL_utf8_foldclosures used below doesn't handle them, so have to be | |
3371 | * handled outside this routine */ | |
3372 | use_chrtest_void = TRUE; | |
3373 | } | |
3374 | else { /* an EXACTFish node which doesn't begin with a multi-char fold */ | |
3375 | c1 = (UTF_PATTERN) ? valid_utf8_to_uvchr(pat, NULL) : *pat; | |
3376 | if (c1 > 256) { | |
3377 | /* Load the folds hash, if not already done */ | |
3378 | SV** listp; | |
3379 | if (! PL_utf8_foldclosures) { | |
3380 | if (! PL_utf8_tofold) { | |
3381 | U8 dummy[UTF8_MAXBYTES+1]; | |
3382 | ||
3383 | /* Force loading this by folding an above-Latin1 char */ | |
3384 | to_utf8_fold((U8*) HYPHEN_UTF8, dummy, NULL); | |
3385 | assert(PL_utf8_tofold); /* Verify that worked */ | |
3386 | } | |
3387 | PL_utf8_foldclosures = _swash_inversion_hash(PL_utf8_tofold); | |
3388 | } | |
3389 | ||
3390 | /* The fold closures data structure is a hash with the keys being | |
3391 | * the UTF-8 of every character that is folded to, like 'k', and | |
3392 | * the values each an array of all code points that fold to its | |
3393 | * key. e.g. [ 'k', 'K', KELVIN_SIGN ]. Multi-character folds are | |
3394 | * not included */ | |
3395 | if ((! (listp = hv_fetch(PL_utf8_foldclosures, | |
3396 | (char *) pat, | |
3397 | UTF8SKIP(pat), | |
3398 | FALSE)))) | |
3399 | { | |
3400 | /* Not found in the hash, therefore there are no folds | |
3401 | * containing it, so there is only a single character that | |
3402 | * could match */ | |
3403 | c2 = c1; | |
3404 | } | |
3405 | else { /* Does participate in folds */ | |
3406 | AV* list = (AV*) *listp; | |
3407 | if (av_len(list) != 1) { | |
3408 | ||
3409 | /* If there aren't exactly two folds to this, it is outside | |
3410 | * the scope of this function */ | |
3411 | use_chrtest_void = TRUE; | |
3412 | } | |
3413 | else { /* There are two. Get them */ | |
3414 | SV** c_p = av_fetch(list, 0, FALSE); | |
3415 | if (c_p == NULL) { | |
3416 | Perl_croak(aTHX_ "panic: invalid PL_utf8_foldclosures structure"); | |
3417 | } | |
3418 | c1 = SvUV(*c_p); | |
3419 | ||
3420 | c_p = av_fetch(list, 1, FALSE); | |
3421 | if (c_p == NULL) { | |
3422 | Perl_croak(aTHX_ "panic: invalid PL_utf8_foldclosures structure"); | |
3423 | } | |
3424 | c2 = SvUV(*c_p); | |
3425 | ||
3426 | /* Folds that cross the 255/256 boundary are forbidden if | |
3427 | * EXACTFL, or EXACTFA and one is ASCIII. Since the | |
3428 | * pattern character is above 256, and its only other match | |
3429 | * is below 256, the only legal match will be to itself. | |
3430 | * We have thrown away the original, so have to compute | |
3431 | * which is the one above 255 */ | |
3432 | if ((c1 < 256) != (c2 < 256)) { | |
3433 | if (OP(text_node) == EXACTFL | |
3434 | || (OP(text_node) == EXACTFA | |
3435 | && (isASCII(c1) || isASCII(c2)))) | |
3436 | { | |
3437 | if (c1 < 256) { | |
3438 | c1 = c2; | |
3439 | } | |
3440 | else { | |
3441 | c2 = c1; | |
3442 | } | |
3443 | } | |
3444 | } | |
3445 | } | |
3446 | } | |
3447 | } | |
3448 | else /* Here, c1 is < 255 */ | |
3449 | if (utf8_target | |
3450 | && HAS_NONLATIN1_FOLD_CLOSURE(c1) | |
3451 | && OP(text_node) != EXACTFL | |
3452 | && (OP(text_node) != EXACTFA || ! isASCII(c1))) | |
c74f6de9 KW |
3453 | { |
3454 | /* Here, there could be something above Latin1 in the target which | |
79a2a0e8 KW |
3455 | * folds to this character in the pattern. All such cases except |
3456 | * LATIN SMALL LETTER Y WITH DIAERESIS have more than two characters | |
3457 | * involved in their folds, so are outside the scope of this | |
3458 | * function */ | |
3459 | if (UNLIKELY(c1 == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) { | |
3460 | c2 = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS; | |
3461 | } | |
3462 | else { | |
3463 | use_chrtest_void = TRUE; | |
3464 | } | |
c74f6de9 KW |
3465 | } |
3466 | else { /* Here nothing above Latin1 can fold to the pattern character */ | |
3467 | switch (OP(text_node)) { | |
3468 | ||
3469 | case EXACTFL: /* /l rules */ | |
79a2a0e8 | 3470 | c2 = PL_fold_locale[c1]; |
c74f6de9 KW |
3471 | break; |
3472 | ||
3473 | case EXACTF: | |
3474 | if (! utf8_target) { /* /d rules */ | |
79a2a0e8 | 3475 | c2 = PL_fold[c1]; |
c74f6de9 KW |
3476 | break; |
3477 | } | |
3478 | /* FALLTHROUGH */ | |
3479 | /* /u rules for all these. This happens to work for | |
79a2a0e8 | 3480 | * EXACTFA as nothing in Latin1 folds to ASCII */ |
c74f6de9 KW |
3481 | case EXACTFA: |
3482 | case EXACTFU_TRICKYFOLD: | |
79a2a0e8 | 3483 | case EXACTFU_SS: |
c74f6de9 | 3484 | case EXACTFU: |
79a2a0e8 | 3485 | c2 = PL_fold_latin1[c1]; |
c74f6de9 KW |
3486 | break; |
3487 | ||
3488 | default: Perl_croak(aTHX_ "panic: Unexpected op %u", OP(text_node)); | |
3489 | } | |
3490 | } | |
3491 | } | |
79a2a0e8 KW |
3492 | |
3493 | /* Here have figured things out. Set up the returns */ | |
3494 | if (use_chrtest_void) { | |
3495 | *c2p = *c1p = CHRTEST_VOID; | |
3496 | } | |
3497 | else if (utf8_target) { | |
3498 | if (! utf8_has_been_setup) { /* Don't have the utf8; must get it */ | |
3499 | uvchr_to_utf8(c1_utf8, c1); | |
3500 | uvchr_to_utf8(c2_utf8, c2); | |
c74f6de9 | 3501 | } |
c74f6de9 | 3502 | |
79a2a0e8 KW |
3503 | /* Invariants are stored in both the utf8 and byte outputs; Use |
3504 | * negative numbers otherwise for the byte ones. Make sure that the | |
3505 | * byte ones are the same iff the utf8 ones are the same */ | |
3506 | *c1p = (UTF8_IS_INVARIANT(*c1_utf8)) ? *c1_utf8 : CHRTEST_NOT_A_CP_1; | |
3507 | *c2p = (UTF8_IS_INVARIANT(*c2_utf8)) | |
3508 | ? *c2_utf8 | |
3509 | : (c1 == c2) | |
3510 | ? CHRTEST_NOT_A_CP_1 | |
3511 | : CHRTEST_NOT_A_CP_2; | |
3512 | } | |
3513 | else if (c1 > 255) { | |
3514 | if (c2 > 255) { /* both possibilities are above what a non-utf8 string | |
3515 | can represent */ | |
3516 | return FALSE; | |
3517 | } | |
c74f6de9 | 3518 | |
79a2a0e8 KW |
3519 | *c1p = *c2p = c2; /* c2 is the only representable value */ |
3520 | } | |
3521 | else { /* c1 is representable; see about c2 */ | |
3522 | *c1p = c1; | |
3523 | *c2p = (c2 < 256) ? c2 : c1; | |
c74f6de9 | 3524 | } |
2f554ef7 | 3525 | |
c74f6de9 KW |
3526 | return TRUE; |
3527 | } | |
2f554ef7 | 3528 | |
f73aaa43 DM |
3529 | /* returns -1 on failure, $+[0] on success */ |
3530 | STATIC I32 | |
3531 | S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) | |
d6a28714 | 3532 | { |
a35a87e7 | 3533 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
3534 | dMY_CXT; |
3535 | #endif | |
27da23d5 | 3536 | dVAR; |
eb578fdb | 3537 | const bool utf8_target = PL_reg_match_utf8; |
4ad0818d | 3538 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
288b8c02 NC |
3539 | REGEXP *rex_sv = reginfo->prog; |
3540 | regexp *rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 3541 | RXi_GET_DECL(rex,rexi); |
2f554ef7 | 3542 | I32 oldsave; |
5d9a96ca | 3543 | /* the current state. This is a cached copy of PL_regmatch_state */ |
eb578fdb | 3544 | regmatch_state *st; |
5d9a96ca | 3545 | /* cache heavy used fields of st in registers */ |
eb578fdb KW |
3546 | regnode *scan; |
3547 | regnode *next; | |
3548 | U32 n = 0; /* general value; init to avoid compiler warning */ | |
3549 | I32 ln = 0; /* len or last; init to avoid compiler warning */ | |
d60de1d1 | 3550 | char *locinput = startpos; |
4d5016e5 | 3551 | char *pushinput; /* where to continue after a PUSH */ |
eb578fdb | 3552 | I32 nextchr; /* is always set to UCHARAT(locinput) */ |
24d3c4a9 | 3553 | |
b69b0499 | 3554 | bool result = 0; /* return value of S_regmatch */ |
24d3c4a9 | 3555 | int depth = 0; /* depth of backtrack stack */ |
4b196cd4 YO |
3556 | U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */ |
3557 | const U32 max_nochange_depth = | |
3558 | (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ? | |
3559 | 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH; | |
77cb431f DM |
3560 | regmatch_state *yes_state = NULL; /* state to pop to on success of |
3561 | subpattern */ | |
e2e6a0f1 YO |
3562 | /* mark_state piggy backs on the yes_state logic so that when we unwind |
3563 | the stack on success we can update the mark_state as we go */ | |
3564 | regmatch_state *mark_state = NULL; /* last mark state we have seen */ | |
faec1544 | 3565 | regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ |
b8591aee | 3566 | struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ |
40a82448 | 3567 | U32 state_num; |
5d458dd8 YO |
3568 | bool no_final = 0; /* prevent failure from backtracking? */ |
3569 | bool do_cutgroup = 0; /* no_final only until next branch/trie entry */ | |
d60de1d1 | 3570 | char *startpoint = locinput; |
5d458dd8 YO |
3571 | SV *popmark = NULL; /* are we looking for a mark? */ |
3572 | SV *sv_commit = NULL; /* last mark name seen in failure */ | |
3573 | SV *sv_yes_mark = NULL; /* last mark name we have seen | |
486ec47a | 3574 | during a successful match */ |
5d458dd8 YO |
3575 | U32 lastopen = 0; /* last open we saw */ |
3576 | bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0; | |
19b95bf0 | 3577 | SV* const oreplsv = GvSV(PL_replgv); |
24d3c4a9 DM |
3578 | /* these three flags are set by various ops to signal information to |
3579 | * the very next op. They have a useful lifetime of exactly one loop | |
3580 | * iteration, and are not preserved or restored by state pushes/pops | |
3581 | */ | |
3582 | bool sw = 0; /* the condition value in (?(cond)a|b) */ | |
3583 | bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */ | |
3584 | int logical = 0; /* the following EVAL is: | |
3585 | 0: (?{...}) | |
3586 | 1: (?(?{...})X|Y) | |
3587 | 2: (??{...}) | |
3588 | or the following IFMATCH/UNLESSM is: | |
3589 | false: plain (?=foo) | |
3590 | true: used as a condition: (?(?=foo)) | |
3591 | */ | |
81ed78b2 DM |
3592 | PAD* last_pad = NULL; |
3593 | dMULTICALL; | |
3594 | I32 gimme = G_SCALAR; | |
3595 | CV *caller_cv = NULL; /* who called us */ | |
3596 | CV *last_pushed_cv = NULL; /* most recently called (?{}) CV */ | |
74088413 | 3597 | CHECKPOINT runops_cp; /* savestack position before executing EVAL */ |
81ed78b2 | 3598 | |
95b24440 | 3599 | #ifdef DEBUGGING |
e68ec53f | 3600 | GET_RE_DEBUG_FLAGS_DECL; |
d6a28714 JH |
3601 | #endif |
3602 | ||
81ed78b2 DM |
3603 | /* shut up 'may be used uninitialized' compiler warnings for dMULTICALL */ |
3604 | multicall_oldcatch = 0; | |
3605 | multicall_cv = NULL; | |
3606 | cx = NULL; | |
4f8dbb2d JL |
3607 | PERL_UNUSED_VAR(multicall_cop); |
3608 | PERL_UNUSED_VAR(newsp); | |
81ed78b2 DM |
3609 | |
3610 | ||
7918f24d NC |
3611 | PERL_ARGS_ASSERT_REGMATCH; |
3612 | ||
3b57cd43 | 3613 | DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({ |
24b23f37 | 3614 | PerlIO_printf(Perl_debug_log,"regmatch start\n"); |
3b57cd43 | 3615 | })); |
5d9a96ca DM |
3616 | /* on first ever call to regmatch, allocate first slab */ |
3617 | if (!PL_regmatch_slab) { | |
3618 | Newx(PL_regmatch_slab, 1, regmatch_slab); | |
3619 | PL_regmatch_slab->prev = NULL; | |
3620 | PL_regmatch_slab->next = NULL; | |
86545054 | 3621 | PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); |
5d9a96ca DM |
3622 | } |
3623 | ||
2f554ef7 DM |
3624 | oldsave = PL_savestack_ix; |
3625 | SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL); | |
3626 | SAVEVPTR(PL_regmatch_slab); | |
3627 | SAVEVPTR(PL_regmatch_state); | |
5d9a96ca DM |
3628 | |
3629 | /* grab next free state slot */ | |
3630 | st = ++PL_regmatch_state; | |
86545054 | 3631 | if (st > SLAB_LAST(PL_regmatch_slab)) |
5d9a96ca DM |
3632 | st = PL_regmatch_state = S_push_slab(aTHX); |
3633 | ||
d6a28714 | 3634 | /* Note that nextchr is a byte even in UTF */ |
7016d6eb | 3635 | SET_nextchr; |
d6a28714 JH |
3636 | scan = prog; |
3637 | while (scan != NULL) { | |
8ba1375e | 3638 | |
a3621e74 | 3639 | DEBUG_EXECUTE_r( { |
6136c704 | 3640 | SV * const prop = sv_newmortal(); |
1de06328 | 3641 | regnode *rnext=regnext(scan); |
f2ed9b32 | 3642 | DUMP_EXEC_POS( locinput, scan, utf8_target ); |
32fc9b6a | 3643 | regprop(rex, prop, scan); |
07be1b83 YO |
3644 | |
3645 | PerlIO_printf(Perl_debug_log, | |
3646 | "%3"IVdf":%*s%s(%"IVdf")\n", | |
f8fc2ecf | 3647 | (IV)(scan - rexi->program), depth*2, "", |
07be1b83 | 3648 | SvPVX_const(prop), |
1de06328 | 3649 | (PL_regkind[OP(scan)] == END || !rnext) ? |
f8fc2ecf | 3650 | 0 : (IV)(rnext - rexi->program)); |
2a782b5b | 3651 | }); |
d6a28714 JH |
3652 | |
3653 | next = scan + NEXT_OFF(scan); | |
3654 | if (next == scan) | |
3655 | next = NULL; | |
40a82448 | 3656 | state_num = OP(scan); |
d6a28714 | 3657 | |
40a82448 | 3658 | reenter_switch: |
34a81e2b | 3659 | |
7016d6eb | 3660 | SET_nextchr; |
e6ca698c | 3661 | assert(nextchr < 256 && (nextchr >= 0 || nextchr == NEXTCHR_EOS)); |
bf798dc4 | 3662 | |
40a82448 | 3663 | switch (state_num) { |
3c0563b9 | 3664 | case BOL: /* /^../ */ |
7fba1cd6 | 3665 | if (locinput == PL_bostr) |
d6a28714 | 3666 | { |
3b0527fe | 3667 | /* reginfo->till = reginfo->bol; */ |
b8c5462f JH |
3668 | break; |
3669 | } | |
d6a28714 | 3670 | sayNO; |
3c0563b9 DM |
3671 | |
3672 | case MBOL: /* /^../m */ | |
12d33761 | 3673 | if (locinput == PL_bostr || |
7016d6eb | 3674 | (!NEXTCHR_IS_EOS && locinput[-1] == '\n')) |
d6a28714 | 3675 | { |
b8c5462f JH |
3676 | break; |
3677 | } | |
d6a28714 | 3678 | sayNO; |
3c0563b9 DM |
3679 | |
3680 | case SBOL: /* /^../s */ | |
c2a73568 | 3681 | if (locinput == PL_bostr) |
b8c5462f | 3682 | break; |
d6a28714 | 3683 | sayNO; |
3c0563b9 DM |
3684 | |
3685 | case GPOS: /* \G */ | |
3b0527fe | 3686 | if (locinput == reginfo->ganch) |
d6a28714 JH |
3687 | break; |
3688 | sayNO; | |
ee9b8eae | 3689 | |
3c0563b9 | 3690 | case KEEPS: /* \K */ |
ee9b8eae | 3691 | /* update the startpoint */ |
b93070ed | 3692 | st->u.keeper.val = rex->offs[0].start; |
b93070ed | 3693 | rex->offs[0].start = locinput - PL_bostr; |
4d5016e5 | 3694 | PUSH_STATE_GOTO(KEEPS_next, next, locinput); |
ee9b8eae YO |
3695 | /*NOT-REACHED*/ |
3696 | case KEEPS_next_fail: | |
3697 | /* rollback the start point change */ | |
b93070ed | 3698 | rex->offs[0].start = st->u.keeper.val; |
ee9b8eae YO |
3699 | sayNO_SILENT; |
3700 | /*NOT-REACHED*/ | |
3c0563b9 DM |
3701 | |
3702 | case EOL: /* /..$/ */ | |
d6a28714 | 3703 | goto seol; |
3c0563b9 DM |
3704 | |
3705 | case MEOL: /* /..$/m */ | |
7016d6eb | 3706 | if (!NEXTCHR_IS_EOS && nextchr != '\n') |
b8c5462f | 3707 | sayNO; |
b8c5462f | 3708 | break; |
3c0563b9 DM |
3709 | |
3710 | case SEOL: /* /..$/s */ | |
d6a28714 | 3711 | seol: |
7016d6eb | 3712 | if (!NEXTCHR_IS_EOS && nextchr != '\n') |
b8c5462f | 3713 | sayNO; |
d6a28714 | 3714 | if (PL_regeol - locinput > 1) |
b8c5462f | 3715 | sayNO; |
b8c5462f | 3716 | break; |
3c0563b9 DM |
3717 | |
3718 | case EOS: /* \z */ | |
7016d6eb | 3719 | if (!NEXTCHR_IS_EOS) |
b8c5462f | 3720 | sayNO; |
d6a28714 | 3721 | break; |
3c0563b9 DM |
3722 | |
3723 | case SANY: /* /./s */ | |
7016d6eb | 3724 | if (NEXTCHR_IS_EOS) |
4633a7c4 | 3725 | sayNO; |
28b98f76 | 3726 | goto increment_locinput; |
3c0563b9 DM |
3727 | |
3728 | case CANY: /* \C */ | |
7016d6eb | 3729 | if (NEXTCHR_IS_EOS) |
f33976b4 | 3730 | sayNO; |
3640db6b | 3731 | locinput++; |
a0d0e21e | 3732 | break; |
3c0563b9 DM |
3733 | |
3734 | case REG_ANY: /* /./ */ | |
7016d6eb | 3735 | if ((NEXTCHR_IS_EOS) || nextchr == '\n') |
1aa99e6b | 3736 | sayNO; |
28b98f76 DM |
3737 | goto increment_locinput; |
3738 | ||
166ba7cd DM |
3739 | |
3740 | #undef ST | |
3741 | #define ST st->u.trie | |
3c0563b9 | 3742 | case TRIEC: /* (ab|cd) with known charclass */ |
786e8c11 YO |
3743 | /* In this case the charclass data is available inline so |
3744 | we can fail fast without a lot of extra overhead. | |
3745 | */ | |
7016d6eb | 3746 | if(!NEXTCHR_IS_EOS && !ANYOF_BITMAP_TEST(scan, nextchr)) { |
fab2782b YO |
3747 | DEBUG_EXECUTE_r( |
3748 | PerlIO_printf(Perl_debug_log, | |
3749 | "%*s %sfailed to match trie start class...%s\n", | |
3750 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) | |
3751 | ); | |
3752 | sayNO_SILENT; | |
118e2215 | 3753 | assert(0); /* NOTREACHED */ |
786e8c11 YO |
3754 | } |
3755 | /* FALL THROUGH */ | |
3c0563b9 | 3756 | case TRIE: /* (ab|cd) */ |
2e64971a DM |
3757 | /* the basic plan of execution of the trie is: |
3758 | * At the beginning, run though all the states, and | |
3759 | * find the longest-matching word. Also remember the position | |
3760 | * of the shortest matching word. For example, this pattern: | |
3761 | * 1 2 3 4 5 | |
3762 | * ab|a|x|abcd|abc | |
3763 | * when matched against the string "abcde", will generate | |
3764 | * accept states for all words except 3, with the longest | |
895cc420 | 3765 | * matching word being 4, and the shortest being 2 (with |
2e64971a DM |
3766 | * the position being after char 1 of the string). |
3767 | * | |
3768 | * Then for each matching word, in word order (i.e. 1,2,4,5), | |
3769 | * we run the remainder of the pattern; on each try setting | |
3770 | * the current position to the character following the word, | |
3771 | * returning to try the next word on failure. | |
3772 | * | |
3773 | * We avoid having to build a list of words at runtime by | |
3774 | * using a compile-time structure, wordinfo[].prev, which | |
3775 | * gives, for each word, the previous accepting word (if any). | |
3776 | * In the case above it would contain the mappings 1->2, 2->0, | |
3777 | * 3->0, 4->5, 5->1. We can use this table to generate, from | |
3778 | * the longest word (4 above), a list of all words, by | |
3779 | * following the list of prev pointers; this gives us the | |
3780 | * unordered list 4,5,1,2. Then given the current word we have | |
3781 | * just tried, we can go through the list and find the | |
3782 | * next-biggest word to try (so if we just failed on word 2, | |
3783 | * the next in the list is 4). | |
3784 | * | |
3785 | * Since at runtime we don't record the matching position in | |
3786 | * the string for each word, we have to work that out for | |
3787 | * each word we're about to process. The wordinfo table holds | |
3788 | * the character length of each word; given that we recorded | |
3789 | * at the start: the position of the shortest word and its | |
3790 | * length in chars, we just need to move the pointer the | |
3791 | * difference between the two char lengths. Depending on | |
3792 | * Unicode status and folding, that's cheap or expensive. | |
3793 | * | |
3794 | * This algorithm is optimised for the case where are only a | |
3795 | * small number of accept states, i.e. 0,1, or maybe 2. | |
3796 | * With lots of accepts states, and having to try all of them, | |
3797 | * it becomes quadratic on number of accept states to find all | |
3798 | * the next words. | |
3799 | */ | |
3800 | ||
3dab1dad | 3801 | { |
07be1b83 | 3802 | /* what type of TRIE am I? (utf8 makes this contextual) */ |
a0a388a1 | 3803 | DECL_TRIE_TYPE(scan); |
3dab1dad YO |
3804 | |
3805 | /* what trie are we using right now */ | |
be8e71aa | 3806 | reg_trie_data * const trie |
f8fc2ecf | 3807 | = (reg_trie_data*)rexi->data->data[ ARG( scan ) ]; |
85fbaab2 | 3808 | HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]); |
3dab1dad | 3809 | U32 state = trie->startstate; |
166ba7cd | 3810 | |
7016d6eb DM |
3811 | if ( trie->bitmap |
3812 | && (NEXTCHR_IS_EOS || !TRIE_BITMAP_TEST(trie, nextchr))) | |
3813 | { | |
3dab1dad YO |
3814 | if (trie->states[ state ].wordnum) { |
3815 | DEBUG_EXECUTE_r( | |
3816 | PerlIO_printf(Perl_debug_log, | |
3817 | "%*s %smatched empty string...%s\n", | |
5bc10b2c | 3818 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad | 3819 | ); |
20dbff7c YO |
3820 | if (!trie->jump) |
3821 | break; | |
3dab1dad YO |
3822 | } else { |
3823 | DEBUG_EXECUTE_r( | |
3824 | PerlIO_printf(Perl_debug_log, | |
786e8c11 | 3825 | "%*s %sfailed to match trie start class...%s\n", |
5bc10b2c | 3826 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad YO |
3827 | ); |
3828 | sayNO_SILENT; | |
3829 | } | |
3830 | } | |
166ba7cd | 3831 | |
786e8c11 YO |
3832 | { |
3833 | U8 *uc = ( U8* )locinput; | |
3834 | ||
3835 | STRLEN len = 0; | |
3836 | STRLEN foldlen = 0; | |
3837 | U8 *uscan = (U8*)NULL; | |
786e8c11 | 3838 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
2e64971a DM |
3839 | U32 charcount = 0; /* how many input chars we have matched */ |
3840 | U32 accepted = 0; /* have we seen any accepting states? */ | |
786e8c11 | 3841 | |
786e8c11 | 3842 | ST.jump = trie->jump; |
786e8c11 | 3843 | ST.me = scan; |
2e64971a DM |
3844 | ST.firstpos = NULL; |
3845 | ST.longfold = FALSE; /* char longer if folded => it's harder */ | |
3846 | ST.nextword = 0; | |
3847 | ||
3848 | /* fully traverse the TRIE; note the position of the | |
3849 | shortest accept state and the wordnum of the longest | |
3850 | accept state */ | |
07be1b83 | 3851 | |
a3621e74 | 3852 | while ( state && uc <= (U8*)PL_regeol ) { |
786e8c11 | 3853 | U32 base = trie->states[ state ].trans.base; |
f9f4320a | 3854 | UV uvc = 0; |
acb909b4 | 3855 | U16 charid = 0; |
2e64971a DM |
3856 | U16 wordnum; |
3857 | wordnum = trie->states[ state ].wordnum; | |
3858 | ||
3859 | if (wordnum) { /* it's an accept state */ | |
3860 | if (!accepted) { | |
3861 | accepted = 1; | |
3862 | /* record first match position */ | |
3863 | if (ST.longfold) { | |
3864 | ST.firstpos = (U8*)locinput; | |
3865 | ST.firstchars = 0; | |
5b47454d | 3866 | } |
2e64971a DM |
3867 | else { |
3868 | ST.firstpos = uc; | |
3869 | ST.firstchars = charcount; | |
3870 | } | |
3871 | } | |
3872 | if (!ST.nextword || wordnum < ST.nextword) | |
3873 | ST.nextword = wordnum; | |
3874 | ST.topword = wordnum; | |
786e8c11 | 3875 | } |
a3621e74 | 3876 | |
07be1b83 | 3877 | DEBUG_TRIE_EXECUTE_r({ |
f2ed9b32 | 3878 | DUMP_EXEC_POS( (char *)uc, scan, utf8_target ); |
a3621e74 | 3879 | PerlIO_printf( Perl_debug_log, |
2e64971a | 3880 | "%*s %sState: %4"UVxf" Accepted: %c ", |
5bc10b2c | 3881 | 2+depth * 2, "", PL_colors[4], |
2e64971a | 3882 | (UV)state, (accepted ? 'Y' : 'N')); |
07be1b83 | 3883 | }); |
a3621e74 | 3884 | |
2e64971a | 3885 | /* read a char and goto next state */ |
7016d6eb | 3886 | if ( base && (foldlen || uc < (U8*)PL_regeol)) { |
6dd2be57 | 3887 | I32 offset; |
55eed653 NC |
3888 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
3889 | uscan, len, uvc, charid, foldlen, | |
3890 | foldbuf, uniflags); | |
2e64971a DM |
3891 | charcount++; |
3892 | if (foldlen>0) | |
3893 | ST.longfold = TRUE; | |
5b47454d | 3894 | if (charid && |
6dd2be57 DM |
3895 | ( ((offset = |
3896 | base + charid - 1 - trie->uniquecharcount)) >= 0) | |
3897 | ||
3898 | && ((U32)offset < trie->lasttrans) | |
3899 | && trie->trans[offset].check == state) | |
5b47454d | 3900 | { |
6dd2be57 | 3901 | state = trie->trans[offset].next; |
5b47454d DM |
3902 | } |
3903 | else { | |
3904 | state = 0; | |
3905 | } | |
3906 | uc += len; | |
3907 | ||
3908 | } | |
3909 | else { | |
a3621e74 YO |
3910 | state = 0; |
3911 | } | |
3912 | DEBUG_TRIE_EXECUTE_r( | |
e4584336 | 3913 | PerlIO_printf( Perl_debug_log, |
786e8c11 | 3914 | "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n", |
e4584336 | 3915 | charid, uvc, (UV)state, PL_colors[5] ); |
a3621e74 YO |
3916 | ); |
3917 | } | |
2e64971a | 3918 | if (!accepted) |
a3621e74 | 3919 | sayNO; |
a3621e74 | 3920 | |
2e64971a DM |
3921 | /* calculate total number of accept states */ |
3922 | { | |
3923 | U16 w = ST.topword; | |
3924 | accepted = 0; | |
3925 | while (w) { | |
3926 | w = trie->wordinfo[w].prev; | |
3927 | accepted++; | |
3928 | } | |
3929 | ST.accepted = accepted; | |
3930 | } | |
3931 | ||
166ba7cd DM |
3932 | DEBUG_EXECUTE_r( |
3933 | PerlIO_printf( Perl_debug_log, | |
3934 | "%*s %sgot %"IVdf" possible matches%s\n", | |
5bc10b2c | 3935 | REPORT_CODE_OFF + depth * 2, "", |
166ba7cd DM |
3936 | PL_colors[4], (IV)ST.accepted, PL_colors[5] ); |
3937 | ); | |
2e64971a | 3938 | goto trie_first_try; /* jump into the fail handler */ |
786e8c11 | 3939 | }} |
118e2215 | 3940 | assert(0); /* NOTREACHED */ |
2e64971a DM |
3941 | |
3942 | case TRIE_next_fail: /* we failed - try next alternative */ | |
a059a757 DM |
3943 | { |
3944 | U8 *uc; | |
fae667d5 YO |
3945 | if ( ST.jump) { |
3946 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 | 3947 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); |
fae667d5 | 3948 | } |
2e64971a DM |
3949 | if (!--ST.accepted) { |
3950 | DEBUG_EXECUTE_r({ | |
3951 | PerlIO_printf( Perl_debug_log, | |
3952 | "%*s %sTRIE failed...%s\n", | |
3953 | REPORT_CODE_OFF+depth*2, "", | |
3954 | PL_colors[4], | |
3955 | PL_colors[5] ); | |
3956 | }); | |
3957 | sayNO_SILENT; | |
3958 | } | |
3959 | { | |
3960 | /* Find next-highest word to process. Note that this code | |
3961 | * is O(N^2) per trie run (O(N) per branch), so keep tight */ | |
eb578fdb KW |
3962 | U16 min = 0; |
3963 | U16 word; | |
3964 | U16 const nextword = ST.nextword; | |
3965 | reg_trie_wordinfo * const wordinfo | |
2e64971a DM |
3966 | = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo; |
3967 | for (word=ST.topword; word; word=wordinfo[word].prev) { | |
3968 | if (word > nextword && (!min || word < min)) | |
3969 | min = word; | |
3970 | } | |
3971 | ST.nextword = min; | |
3972 | } | |
3973 | ||
fae667d5 | 3974 | trie_first_try: |
5d458dd8 YO |
3975 | if (do_cutgroup) { |
3976 | do_cutgroup = 0; | |
3977 | no_final = 0; | |
3978 | } | |
fae667d5 YO |
3979 | |
3980 | if ( ST.jump) { | |
b93070ed | 3981 | ST.lastparen = rex->lastparen; |
f6033a9d | 3982 | ST.lastcloseparen = rex->lastcloseparen; |
fae667d5 | 3983 | REGCP_SET(ST.cp); |
2e64971a | 3984 | } |
a3621e74 | 3985 | |
2e64971a | 3986 | /* find start char of end of current word */ |
166ba7cd | 3987 | { |
2e64971a | 3988 | U32 chars; /* how many chars to skip */ |
2e64971a DM |
3989 | reg_trie_data * const trie |
3990 | = (reg_trie_data*)rexi->data->data[ARG(ST.me)]; | |
3991 | ||
3992 | assert((trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3993 | >= ST.firstchars); | |
3994 | chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3995 | - ST.firstchars; | |
a059a757 | 3996 | uc = ST.firstpos; |
2e64971a DM |
3997 | |
3998 | if (ST.longfold) { | |
3999 | /* the hard option - fold each char in turn and find | |
4000 | * its folded length (which may be different */ | |
4001 | U8 foldbuf[UTF8_MAXBYTES_CASE + 1]; | |
4002 | STRLEN foldlen; | |
4003 | STRLEN len; | |
d9a396a3 | 4004 | UV uvc; |
2e64971a DM |
4005 | U8 *uscan; |
4006 | ||
4007 | while (chars) { | |
f2ed9b32 | 4008 | if (utf8_target) { |
2e64971a DM |
4009 | uvc = utf8n_to_uvuni((U8*)uc, UTF8_MAXLEN, &len, |
4010 | uniflags); | |
4011 | uc += len; | |
4012 | } | |
4013 | else { | |
4014 | uvc = *uc; | |
4015 | uc++; | |
4016 | } | |
4017 | uvc = to_uni_fold(uvc, foldbuf, &foldlen); | |
4018 | uscan = foldbuf; | |
4019 | while (foldlen) { | |
4020 | if (!--chars) | |
4021 | break; | |
4022 | uvc = utf8n_to_uvuni(uscan, UTF8_MAXLEN, &len, | |
4023 | uniflags); | |
4024 | uscan += len; | |
4025 | foldlen -= len; | |
4026 | } | |
4027 | } | |
a3621e74 | 4028 | } |
2e64971a | 4029 | else { |
f2ed9b32 | 4030 | if (utf8_target) |
2e64971a DM |
4031 | while (chars--) |
4032 | uc += UTF8SKIP(uc); | |
4033 | else | |
4034 | uc += chars; | |
4035 | } | |
2e64971a | 4036 | } |
166ba7cd | 4037 | |
6603fe3e DM |
4038 | scan = ST.me + ((ST.jump && ST.jump[ST.nextword]) |
4039 | ? ST.jump[ST.nextword] | |
4040 | : NEXT_OFF(ST.me)); | |
166ba7cd | 4041 | |
2e64971a DM |
4042 | DEBUG_EXECUTE_r({ |
4043 | PerlIO_printf( Perl_debug_log, | |
4044 | "%*s %sTRIE matched word #%d, continuing%s\n", | |
4045 | REPORT_CODE_OFF+depth*2, "", | |
4046 | PL_colors[4], | |
4047 | ST.nextword, | |
4048 | PL_colors[5] | |
4049 | ); | |
4050 | }); | |
4051 | ||
4052 | if (ST.accepted > 1 || has_cutgroup) { | |
a059a757 | 4053 | PUSH_STATE_GOTO(TRIE_next, scan, (char*)uc); |
118e2215 | 4054 | assert(0); /* NOTREACHED */ |
166ba7cd | 4055 | } |
2e64971a DM |
4056 | /* only one choice left - just continue */ |
4057 | DEBUG_EXECUTE_r({ | |
4058 | AV *const trie_words | |
4059 | = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]); | |
4060 | SV ** const tmp = av_fetch( trie_words, | |
4061 | ST.nextword-1, 0 ); | |
4062 | SV *sv= tmp ? sv_newmortal() : NULL; | |
4063 | ||
4064 | PerlIO_printf( Perl_debug_log, | |
4065 | "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n", | |
4066 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], | |
4067 | ST.nextword, | |
4068 | tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0, | |
4069 | PL_colors[0], PL_colors[1], | |
c89df6cf | 4070 | (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)|PERL_PV_ESCAPE_NONASCII |
2e64971a DM |
4071 | ) |
4072 | : "not compiled under -Dr", | |
4073 | PL_colors[5] ); | |
4074 | }); | |
4075 | ||
a059a757 | 4076 | locinput = (char*)uc; |
2e64971a | 4077 | continue; /* execute rest of RE */ |
118e2215 | 4078 | assert(0); /* NOTREACHED */ |
a059a757 | 4079 | } |
166ba7cd DM |
4080 | #undef ST |
4081 | ||
3c0563b9 | 4082 | case EXACT: { /* /abc/ */ |
95b24440 | 4083 | char *s = STRING(scan); |
24d3c4a9 | 4084 | ln = STR_LEN(scan); |
f2ed9b32 | 4085 | if (utf8_target != UTF_PATTERN) { |
bc517b45 | 4086 | /* The target and the pattern have differing utf8ness. */ |
1aa99e6b | 4087 | char *l = locinput; |
24d3c4a9 | 4088 | const char * const e = s + ln; |
a72c7584 | 4089 | |
f2ed9b32 | 4090 | if (utf8_target) { |
e6a3850e KW |
4091 | /* The target is utf8, the pattern is not utf8. |
4092 | * Above-Latin1 code points can't match the pattern; | |
4093 | * invariants match exactly, and the other Latin1 ones need | |
4094 | * to be downgraded to a single byte in order to do the | |
4095 | * comparison. (If we could be confident that the target | |
4096 | * is not malformed, this could be refactored to have fewer | |
4097 | * tests by just assuming that if the first bytes match, it | |
4098 | * is an invariant, but there are tests in the test suite | |
4099 | * dealing with (??{...}) which violate this) */ | |
1aa99e6b IH |
4100 | while (s < e) { |
4101 | if (l >= PL_regeol) | |
5ff6fc6d | 4102 | sayNO; |
e6a3850e KW |
4103 | if (UTF8_IS_ABOVE_LATIN1(* (U8*) l)) { |
4104 | sayNO; | |
4105 | } | |
4106 | if (UTF8_IS_INVARIANT(*(U8*)l)) { | |
4107 | if (*l != *s) { | |
4108 | sayNO; | |
4109 | } | |
4110 | l++; | |
4111 | } | |
4112 | else { | |
4113 | if (TWO_BYTE_UTF8_TO_UNI(*l, *(l+1)) != * (U8*) s) { | |
4114 | sayNO; | |
4115 | } | |
4116 | l += 2; | |
4117 | } | |
4118 | s++; | |
1aa99e6b | 4119 | } |
5ff6fc6d JH |
4120 | } |
4121 | else { | |
4122 | /* The target is not utf8, the pattern is utf8. */ | |
1aa99e6b | 4123 | while (s < e) { |
e6a3850e KW |
4124 | if (l >= PL_regeol || UTF8_IS_ABOVE_LATIN1(* (U8*) s)) |
4125 | { | |
4126 | sayNO; | |
4127 | } | |
4128 | if (UTF8_IS_INVARIANT(*(U8*)s)) { | |
4129 | if (*s != *l) { | |
4130 | sayNO; | |
4131 | } | |
4132 | s++; | |
4133 | } | |
4134 | else { | |
4135 | if (TWO_BYTE_UTF8_TO_UNI(*s, *(s+1)) != * (U8*) l) { | |
4136 | sayNO; | |
4137 | } | |
4138 | s += 2; | |
4139 | } | |
4140 | l++; | |
1aa99e6b | 4141 | } |
5ff6fc6d | 4142 | } |
1aa99e6b | 4143 | locinput = l; |
1aa99e6b IH |
4144 | break; |
4145 | } | |
bc517b45 | 4146 | /* The target and the pattern have the same utf8ness. */ |
d6a28714 JH |
4147 | /* Inline the first character, for speed. */ |
4148 | if (UCHARAT(s) != nextchr) | |
4149 | sayNO; | |
24d3c4a9 | 4150 | if (PL_regeol - locinput < ln) |
d6a28714 | 4151 | sayNO; |
24d3c4a9 | 4152 | if (ln > 1 && memNE(s, locinput, ln)) |
d6a28714 | 4153 | sayNO; |
24d3c4a9 | 4154 | locinput += ln; |
d6a28714 | 4155 | break; |
95b24440 | 4156 | } |
7016d6eb | 4157 | |
3c0563b9 | 4158 | case EXACTFL: { /* /abc/il */ |
a932d541 | 4159 | re_fold_t folder; |
9a5a5549 KW |
4160 | const U8 * fold_array; |
4161 | const char * s; | |
d513472c | 4162 | U32 fold_utf8_flags; |
9a5a5549 | 4163 | |
b8c5462f | 4164 | PL_reg_flags |= RF_tainted; |
9a5a5549 KW |
4165 | folder = foldEQ_locale; |
4166 | fold_array = PL_fold_locale; | |
17580e7a | 4167 | fold_utf8_flags = FOLDEQ_UTF8_LOCALE; |
9a5a5549 KW |
4168 | goto do_exactf; |
4169 | ||
3c0563b9 DM |
4170 | case EXACTFU_SS: /* /\x{df}/iu */ |
4171 | case EXACTFU_TRICKYFOLD: /* /\x{390}/iu */ | |
4172 | case EXACTFU: /* /abc/iu */ | |
9a5a5549 KW |
4173 | folder = foldEQ_latin1; |
4174 | fold_array = PL_fold_latin1; | |
2daa8fee | 4175 | fold_utf8_flags = (UTF_PATTERN) ? FOLDEQ_S1_ALREADY_FOLDED : 0; |
9a5a5549 KW |
4176 | goto do_exactf; |
4177 | ||
3c0563b9 | 4178 | case EXACTFA: /* /abc/iaa */ |
2f7f8cb1 KW |
4179 | folder = foldEQ_latin1; |
4180 | fold_array = PL_fold_latin1; | |
57014d77 | 4181 | fold_utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; |
2f7f8cb1 KW |
4182 | goto do_exactf; |
4183 | ||
3c0563b9 | 4184 | case EXACTF: /* /abc/i */ |
9a5a5549 KW |
4185 | folder = foldEQ; |
4186 | fold_array = PL_fold; | |
62bf7766 | 4187 | fold_utf8_flags = 0; |
9a5a5549 KW |
4188 | |
4189 | do_exactf: | |
4190 | s = STRING(scan); | |
24d3c4a9 | 4191 | ln = STR_LEN(scan); |
d6a28714 | 4192 | |
3c760661 KW |
4193 | if (utf8_target || UTF_PATTERN || state_num == EXACTFU_SS) { |
4194 | /* Either target or the pattern are utf8, or has the issue where | |
4195 | * the fold lengths may differ. */ | |
be8e71aa | 4196 | const char * const l = locinput; |
d07ddd77 | 4197 | char *e = PL_regeol; |
bc517b45 | 4198 | |
d513472c | 4199 | if (! foldEQ_utf8_flags(s, 0, ln, cBOOL(UTF_PATTERN), |
fa5b1667 | 4200 | l, &e, 0, utf8_target, fold_utf8_flags)) |
c3e1d013 KW |
4201 | { |
4202 | sayNO; | |
5486206c | 4203 | } |
d07ddd77 | 4204 | locinput = e; |
d07ddd77 | 4205 | break; |
a0ed51b3 | 4206 | } |
d6a28714 | 4207 | |
0a138b74 | 4208 | /* Neither the target nor the pattern are utf8 */ |
1443c94c DM |
4209 | if (UCHARAT(s) != nextchr |
4210 | && !NEXTCHR_IS_EOS | |
4211 | && UCHARAT(s) != fold_array[nextchr]) | |
9a5a5549 | 4212 | { |
a0ed51b3 | 4213 | sayNO; |
9a5a5549 | 4214 | } |
24d3c4a9 | 4215 | if (PL_regeol - locinput < ln) |
b8c5462f | 4216 | sayNO; |
9a5a5549 | 4217 | if (ln > 1 && ! folder(s, locinput, ln)) |
4633a7c4 | 4218 | sayNO; |
24d3c4a9 | 4219 | locinput += ln; |
a0d0e21e | 4220 | break; |
9a5a5549 | 4221 | } |
63ac0dad KW |
4222 | |
4223 | /* XXX Could improve efficiency by separating these all out using a | |
4224 | * macro or in-line function. At that point regcomp.c would no longer | |
4225 | * have to set the FLAGS fields of these */ | |
3c0563b9 DM |
4226 | case BOUNDL: /* /\b/l */ |
4227 | case NBOUNDL: /* /\B/l */ | |
b2680017 YO |
4228 | PL_reg_flags |= RF_tainted; |
4229 | /* FALL THROUGH */ | |
3c0563b9 DM |
4230 | case BOUND: /* /\b/ */ |
4231 | case BOUNDU: /* /\b/u */ | |
4232 | case BOUNDA: /* /\b/a */ | |
4233 | case NBOUND: /* /\B/ */ | |
4234 | case NBOUNDU: /* /\B/u */ | |
4235 | case NBOUNDA: /* /\B/a */ | |
b2680017 | 4236 | /* was last char in word? */ |
f2e96b5d KW |
4237 | if (utf8_target |
4238 | && FLAGS(scan) != REGEX_ASCII_RESTRICTED_CHARSET | |
4239 | && FLAGS(scan) != REGEX_ASCII_MORE_RESTRICTED_CHARSET) | |
4240 | { | |
b2680017 YO |
4241 | if (locinput == PL_bostr) |
4242 | ln = '\n'; | |
4243 | else { | |
4244 | const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); | |
4245 | ||
4246 | ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); | |
4247 | } | |
63ac0dad | 4248 | if (FLAGS(scan) != REGEX_LOCALE_CHARSET) { |
b2680017 | 4249 | ln = isALNUM_uni(ln); |
7016d6eb DM |
4250 | if (NEXTCHR_IS_EOS) |
4251 | n = 0; | |
4252 | else { | |
4253 | LOAD_UTF8_CHARCLASS_ALNUM(); | |
4254 | n = swash_fetch(PL_utf8_alnum, (U8*)locinput, | |
4255 | utf8_target); | |
4256 | } | |
b2680017 YO |
4257 | } |
4258 | else { | |
4259 | ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln)); | |
7016d6eb | 4260 | n = NEXTCHR_IS_EOS ? 0 : isALNUM_LC_utf8((U8*)locinput); |
b2680017 YO |
4261 | } |
4262 | } | |
4263 | else { | |
cfaf538b KW |
4264 | |
4265 | /* Here the string isn't utf8, or is utf8 and only ascii | |
4266 | * characters are to match \w. In the latter case looking at | |
4267 | * the byte just prior to the current one may be just the final | |
4268 | * byte of a multi-byte character. This is ok. There are two | |
4269 | * cases: | |
4270 | * 1) it is a single byte character, and then the test is doing | |
4271 | * just what it's supposed to. | |
4272 | * 2) it is a multi-byte character, in which case the final | |
4273 | * byte is never mistakable for ASCII, and so the test | |
4274 | * will say it is not a word character, which is the | |
4275 | * correct answer. */ | |
b2680017 YO |
4276 | ln = (locinput != PL_bostr) ? |
4277 | UCHARAT(locinput - 1) : '\n'; | |
63ac0dad KW |
4278 | switch (FLAGS(scan)) { |
4279 | case REGEX_UNICODE_CHARSET: | |
4280 | ln = isWORDCHAR_L1(ln); | |
7016d6eb | 4281 | n = NEXTCHR_IS_EOS ? 0 : isWORDCHAR_L1(nextchr); |
63ac0dad KW |
4282 | break; |
4283 | case REGEX_LOCALE_CHARSET: | |
4284 | ln = isALNUM_LC(ln); | |
7016d6eb | 4285 | n = NEXTCHR_IS_EOS ? 0 : isALNUM_LC(nextchr); |
63ac0dad KW |
4286 | break; |
4287 | case REGEX_DEPENDS_CHARSET: | |
4288 | ln = isALNUM(ln); | |
7016d6eb | 4289 | n = NEXTCHR_IS_EOS ? 0 : isALNUM(nextchr); |
63ac0dad | 4290 | break; |
cfaf538b | 4291 | case REGEX_ASCII_RESTRICTED_CHARSET: |
c973bd4f | 4292 | case REGEX_ASCII_MORE_RESTRICTED_CHARSET: |
cfaf538b | 4293 | ln = isWORDCHAR_A(ln); |
7016d6eb | 4294 | n = NEXTCHR_IS_EOS ? 0 : isWORDCHAR_A(nextchr); |
cfaf538b | 4295 | break; |
63ac0dad KW |
4296 | default: |
4297 | Perl_croak(aTHX_ "panic: Unexpected FLAGS %u in op %u", FLAGS(scan), OP(scan)); | |
4298 | break; | |
b2680017 YO |
4299 | } |
4300 | } | |
63ac0dad KW |
4301 | /* Note requires that all BOUNDs be lower than all NBOUNDs in |
4302 | * regcomp.sym */ | |
4303 | if (((!ln) == (!n)) == (OP(scan) < NBOUND)) | |
b2680017 YO |
4304 | sayNO; |
4305 | break; | |
3c0563b9 | 4306 | |
3c0563b9 | 4307 | case ANYOF: /* /[abc]/ */ |
7016d6eb DM |
4308 | if (NEXTCHR_IS_EOS) |
4309 | sayNO; | |
e0193e47 | 4310 | if (utf8_target) { |
9e55ce06 | 4311 | STRLEN inclasslen = PL_regeol - locinput; |
f2ed9b32 | 4312 | if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target)) |
09b08e9b | 4313 | sayNO; |
b32d7d3e | 4314 | locinput += inclasslen; |
e0f9d4a8 | 4315 | break; |
ffc61ed2 JH |
4316 | } |
4317 | else { | |
20ed0b26 | 4318 | if (!REGINCLASS(rex, scan, (U8*)locinput)) |
09b08e9b | 4319 | sayNO; |
3640db6b | 4320 | locinput++; |
e0f9d4a8 JH |
4321 | break; |
4322 | } | |
b8c5462f | 4323 | break; |
3c0563b9 DM |
4324 | |
4325 | /* Special char classes: \d, \w etc. | |
4326 | * The defines start on line 166 or so */ | |
ee9a90b8 KW |
4327 | CCC_TRY_U(ALNUM, NALNUM, isWORDCHAR, |
4328 | ALNUML, NALNUML, isALNUM_LC, isALNUM_LC_utf8, | |
4329 | ALNUMU, NALNUMU, isWORDCHAR_L1, | |
cfaf538b | 4330 | ALNUMA, NALNUMA, isWORDCHAR_A, |
779d7b58 | 4331 | alnum, "a"); |
ee9a90b8 KW |
4332 | |
4333 | CCC_TRY_U(SPACE, NSPACE, isSPACE, | |
4334 | SPACEL, NSPACEL, isSPACE_LC, isSPACE_LC_utf8, | |
4335 | SPACEU, NSPACEU, isSPACE_L1, | |
cfaf538b | 4336 | SPACEA, NSPACEA, isSPACE_A, |
779d7b58 | 4337 | space, " "); |
ee9a90b8 KW |
4338 | |
4339 | CCC_TRY(DIGIT, NDIGIT, isDIGIT, | |
4340 | DIGITL, NDIGITL, isDIGIT_LC, isDIGIT_LC_utf8, | |
cfaf538b | 4341 | DIGITA, NDIGITA, isDIGIT_A, |
779d7b58 | 4342 | digit, "0"); |
20d0b1e9 | 4343 | |
3c0563b9 | 4344 | case POSIXA: /* /[[:ascii:]]/ etc */ |
7016d6eb | 4345 | if (NEXTCHR_IS_EOS || ! _generic_isCC_A(nextchr, FLAGS(scan))) { |
0658cdde KW |
4346 | sayNO; |
4347 | } | |
4348 | /* Matched a utf8-invariant, so don't have to worry about utf8 */ | |
3640db6b | 4349 | locinput++; |
0658cdde | 4350 | break; |
3c0563b9 DM |
4351 | |
4352 | case NPOSIXA: /* /[^[:ascii:]]/ etc */ | |
7016d6eb | 4353 | if (NEXTCHR_IS_EOS || _generic_isCC_A(nextchr, FLAGS(scan))) { |
0658cdde KW |
4354 | sayNO; |
4355 | } | |
28b98f76 | 4356 | goto increment_locinput; |
0658cdde | 4357 | |
37e2e78e KW |
4358 | case CLUMP: /* Match \X: logical Unicode character. This is defined as |
4359 | a Unicode extended Grapheme Cluster */ | |
4360 | /* From http://www.unicode.org/reports/tr29 (5.2 version). An | |
4361 | extended Grapheme Cluster is: | |
4362 | ||
4363 | CR LF | |
4364 | | Prepend* Begin Extend* | |
4365 | | . | |
4366 | ||
1e958ea9 KW |
4367 | Begin is: ( Special_Begin | ! Control ) |
4368 | Special_Begin is: ( Regional-Indicator+ | Hangul-syllable ) | |
4369 | Extend is: ( Grapheme_Extend | Spacing_Mark ) | |
4370 | Control is: [ GCB_Control CR LF ] | |
4371 | Hangul-syllable is: ( T+ | ( L* ( L | ( LVT | ( V | LV ) V* ) T* ) )) | |
37e2e78e | 4372 | |
27d4fc33 KW |
4373 | If we create a 'Regular_Begin' = Begin - Special_Begin, then |
4374 | we can rewrite | |
4375 | ||
4376 | Begin is ( Regular_Begin + Special Begin ) | |
4377 | ||
4378 | It turns out that 98.4% of all Unicode code points match | |
4379 | Regular_Begin. Doing it this way eliminates a table match in | |
c101f46d | 4380 | the previous implementation for almost all Unicode code points. |
27d4fc33 | 4381 | |
37e2e78e KW |
4382 | There is a subtlety with Prepend* which showed up in testing. |
4383 | Note that the Begin, and only the Begin is required in: | |
4384 | | Prepend* Begin Extend* | |
cc3b396d KW |
4385 | Also, Begin contains '! Control'. A Prepend must be a |
4386 | '! Control', which means it must also be a Begin. What it | |
4387 | comes down to is that if we match Prepend* and then find no | |
4388 | suitable Begin afterwards, that if we backtrack the last | |
4389 | Prepend, that one will be a suitable Begin. | |
37e2e78e KW |
4390 | */ |
4391 | ||
7016d6eb | 4392 | if (NEXTCHR_IS_EOS) |
a0ed51b3 | 4393 | sayNO; |
f2ed9b32 | 4394 | if (! utf8_target) { |
37e2e78e KW |
4395 | |
4396 | /* Match either CR LF or '.', as all the other possibilities | |
4397 | * require utf8 */ | |
4398 | locinput++; /* Match the . or CR */ | |
cc3b396d KW |
4399 | if (nextchr == '\r' /* And if it was CR, and the next is LF, |
4400 | match the LF */ | |
37e2e78e KW |
4401 | && locinput < PL_regeol |
4402 | && UCHARAT(locinput) == '\n') locinput++; | |
4403 | } | |
4404 | else { | |
4405 | ||
4406 | /* Utf8: See if is ( CR LF ); already know that locinput < | |
4407 | * PL_regeol, so locinput+1 is in bounds */ | |
7016d6eb DM |
4408 | if ( nextchr == '\r' && locinput+1 < PL_regeol |
4409 | && UCHARAT(locinput + 1) == '\n') | |
4410 | { | |
37e2e78e KW |
4411 | locinput += 2; |
4412 | } | |
4413 | else { | |
45fdf108 KW |
4414 | STRLEN len; |
4415 | ||
37e2e78e KW |
4416 | /* In case have to backtrack to beginning, then match '.' */ |
4417 | char *starting = locinput; | |
4418 | ||
4419 | /* In case have to backtrack the last prepend */ | |
4420 | char *previous_prepend = 0; | |
4421 | ||
4422 | LOAD_UTF8_CHARCLASS_GCB(); | |
4423 | ||
45fdf108 KW |
4424 | /* Match (prepend)* */ |
4425 | while (locinput < PL_regeol | |
4426 | && (len = is_GCB_Prepend_utf8(locinput))) | |
4427 | { | |
4428 | previous_prepend = locinput; | |
4429 | locinput += len; | |
a1853d78 | 4430 | } |
37e2e78e KW |
4431 | |
4432 | /* As noted above, if we matched a prepend character, but | |
4433 | * the next thing won't match, back off the last prepend we | |
4434 | * matched, as it is guaranteed to match the begin */ | |
4435 | if (previous_prepend | |
4436 | && (locinput >= PL_regeol | |
c101f46d KW |
4437 | || (! swash_fetch(PL_utf8_X_regular_begin, |
4438 | (U8*)locinput, utf8_target) | |
45fdf108 | 4439 | && ! is_GCB_SPECIAL_BEGIN_utf8(locinput))) |
c101f46d | 4440 | ) |
37e2e78e KW |
4441 | { |
4442 | locinput = previous_prepend; | |
4443 | } | |
4444 | ||
4445 | /* Note that here we know PL_regeol > locinput, as we | |
4446 | * tested that upon input to this switch case, and if we | |
4447 | * moved locinput forward, we tested the result just above | |
4448 | * and it either passed, or we backed off so that it will | |
4449 | * now pass */ | |
11dfcd49 KW |
4450 | if (swash_fetch(PL_utf8_X_regular_begin, |
4451 | (U8*)locinput, utf8_target)) { | |
27d4fc33 KW |
4452 | locinput += UTF8SKIP(locinput); |
4453 | } | |
45fdf108 | 4454 | else if (! is_GCB_SPECIAL_BEGIN_utf8(locinput)) { |
37e2e78e KW |
4455 | |
4456 | /* Here did not match the required 'Begin' in the | |
4457 | * second term. So just match the very first | |
4458 | * character, the '.' of the final term of the regex */ | |
4459 | locinput = starting + UTF8SKIP(starting); | |
27d4fc33 | 4460 | goto exit_utf8; |
37e2e78e KW |
4461 | } else { |
4462 | ||
11dfcd49 KW |
4463 | /* Here is a special begin. It can be composed of |
4464 | * several individual characters. One possibility is | |
4465 | * RI+ */ | |
45fdf108 KW |
4466 | if ((len = is_GCB_RI_utf8(locinput))) { |
4467 | locinput += len; | |
11dfcd49 | 4468 | while (locinput < PL_regeol |
45fdf108 | 4469 | && (len = is_GCB_RI_utf8(locinput))) |
11dfcd49 | 4470 | { |
45fdf108 | 4471 | locinput += len; |
11dfcd49 | 4472 | } |
45fdf108 KW |
4473 | } else if ((len = is_GCB_T_utf8(locinput))) { |
4474 | /* Another possibility is T+ */ | |
4475 | locinput += len; | |
11dfcd49 | 4476 | while (locinput < PL_regeol |
45fdf108 | 4477 | && (len = is_GCB_T_utf8(locinput))) |
11dfcd49 | 4478 | { |
45fdf108 | 4479 | locinput += len; |
11dfcd49 KW |
4480 | } |
4481 | } else { | |
4482 | ||
4483 | /* Here, neither RI+ nor T+; must be some other | |
4484 | * Hangul. That means it is one of the others: L, | |
4485 | * LV, LVT or V, and matches: | |
4486 | * L* (L | LVT T* | V * V* T* | LV V* T*) */ | |
4487 | ||
4488 | /* Match L* */ | |
4489 | while (locinput < PL_regeol | |
45fdf108 | 4490 | && (len = is_GCB_L_utf8(locinput))) |
11dfcd49 | 4491 | { |
45fdf108 | 4492 | locinput += len; |
11dfcd49 | 4493 | } |
37e2e78e | 4494 | |
11dfcd49 KW |
4495 | /* Here, have exhausted L*. If the next character |
4496 | * is not an LV, LVT nor V, it means we had to have | |
4497 | * at least one L, so matches L+ in the original | |
4498 | * equation, we have a complete hangul syllable. | |
4499 | * Are done. */ | |
4500 | ||
4501 | if (locinput < PL_regeol | |
45fdf108 | 4502 | && is_GCB_LV_LVT_V_utf8(locinput)) |
11dfcd49 KW |
4503 | { |
4504 | ||
4505 | /* Otherwise keep going. Must be LV, LVT or V. | |
4506 | * See if LVT */ | |
4507 | if (is_utf8_X_LVT((U8*)locinput)) { | |
4508 | locinput += UTF8SKIP(locinput); | |
4509 | } else { | |
4510 | ||
4511 | /* Must be V or LV. Take it, then match | |
4512 | * V* */ | |
4513 | locinput += UTF8SKIP(locinput); | |
4514 | while (locinput < PL_regeol | |
45fdf108 | 4515 | && (len = is_GCB_V_utf8(locinput))) |
11dfcd49 | 4516 | { |
45fdf108 | 4517 | locinput += len; |
11dfcd49 KW |
4518 | } |
4519 | } | |
37e2e78e | 4520 | |
11dfcd49 | 4521 | /* And any of LV, LVT, or V can be followed |
45fdf108 | 4522 | * by T* */ |
11dfcd49 | 4523 | while (locinput < PL_regeol |
45fdf108 | 4524 | && (len = is_GCB_T_utf8(locinput))) |
11dfcd49 | 4525 | { |
45fdf108 | 4526 | locinput += len; |
11dfcd49 KW |
4527 | } |
4528 | } | |
cd94d768 | 4529 | } |
11dfcd49 | 4530 | } |
37e2e78e | 4531 | |
11dfcd49 KW |
4532 | /* Match any extender */ |
4533 | while (locinput < PL_regeol | |
4534 | && swash_fetch(PL_utf8_X_extend, | |
4535 | (U8*)locinput, utf8_target)) | |
4536 | { | |
4537 | locinput += UTF8SKIP(locinput); | |
4538 | } | |
37e2e78e | 4539 | } |
27d4fc33 | 4540 | exit_utf8: |
37e2e78e KW |
4541 | if (locinput > PL_regeol) sayNO; |
4542 | } | |
a0ed51b3 | 4543 | break; |
81714fb9 | 4544 | |
3c0563b9 | 4545 | case NREFFL: /* /\g{name}/il */ |
d7ef4b73 KW |
4546 | { /* The capture buffer cases. The ones beginning with N for the |
4547 | named buffers just convert to the equivalent numbered and | |
4548 | pretend they were called as the corresponding numbered buffer | |
4549 | op. */ | |
26ecd678 TC |
4550 | /* don't initialize these in the declaration, it makes C++ |
4551 | unhappy */ | |
81714fb9 | 4552 | char *s; |
ff1157ca | 4553 | char type; |
8368298a TC |
4554 | re_fold_t folder; |
4555 | const U8 *fold_array; | |
26ecd678 | 4556 | UV utf8_fold_flags; |
8368298a | 4557 | |
81714fb9 | 4558 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
4559 | folder = foldEQ_locale; |
4560 | fold_array = PL_fold_locale; | |
4561 | type = REFFL; | |
17580e7a | 4562 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4563 | goto do_nref; |
4564 | ||
3c0563b9 | 4565 | case NREFFA: /* /\g{name}/iaa */ |
2f7f8cb1 KW |
4566 | folder = foldEQ_latin1; |
4567 | fold_array = PL_fold_latin1; | |
4568 | type = REFFA; | |
4569 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4570 | goto do_nref; | |
4571 | ||
3c0563b9 | 4572 | case NREFFU: /* /\g{name}/iu */ |
d7ef4b73 KW |
4573 | folder = foldEQ_latin1; |
4574 | fold_array = PL_fold_latin1; | |
4575 | type = REFFU; | |
d513472c | 4576 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4577 | goto do_nref; |
4578 | ||
3c0563b9 | 4579 | case NREFF: /* /\g{name}/i */ |
d7ef4b73 KW |
4580 | folder = foldEQ; |
4581 | fold_array = PL_fold; | |
4582 | type = REFF; | |
d513472c | 4583 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4584 | goto do_nref; |
4585 | ||
3c0563b9 | 4586 | case NREF: /* /\g{name}/ */ |
d7ef4b73 | 4587 | type = REF; |
83d7b90b KW |
4588 | folder = NULL; |
4589 | fold_array = NULL; | |
d513472c | 4590 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4591 | do_nref: |
4592 | ||
4593 | /* For the named back references, find the corresponding buffer | |
4594 | * number */ | |
0a4db386 YO |
4595 | n = reg_check_named_buff_matched(rex,scan); |
4596 | ||
d7ef4b73 | 4597 | if ( ! n ) { |
81714fb9 | 4598 | sayNO; |
d7ef4b73 KW |
4599 | } |
4600 | goto do_nref_ref_common; | |
4601 | ||
3c0563b9 | 4602 | case REFFL: /* /\1/il */ |
3280af22 | 4603 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
4604 | folder = foldEQ_locale; |
4605 | fold_array = PL_fold_locale; | |
17580e7a | 4606 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4607 | goto do_ref; |
4608 | ||
3c0563b9 | 4609 | case REFFA: /* /\1/iaa */ |
2f7f8cb1 KW |
4610 | folder = foldEQ_latin1; |
4611 | fold_array = PL_fold_latin1; | |
4612 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4613 | goto do_ref; | |
4614 | ||
3c0563b9 | 4615 | case REFFU: /* /\1/iu */ |
d7ef4b73 KW |
4616 | folder = foldEQ_latin1; |
4617 | fold_array = PL_fold_latin1; | |
d513472c | 4618 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4619 | goto do_ref; |
4620 | ||
3c0563b9 | 4621 | case REFF: /* /\1/i */ |
d7ef4b73 KW |
4622 | folder = foldEQ; |
4623 | fold_array = PL_fold; | |
d513472c | 4624 | utf8_fold_flags = 0; |
83d7b90b | 4625 | goto do_ref; |
d7ef4b73 | 4626 | |
3c0563b9 | 4627 | case REF: /* /\1/ */ |
83d7b90b KW |
4628 | folder = NULL; |
4629 | fold_array = NULL; | |
d513472c | 4630 | utf8_fold_flags = 0; |
83d7b90b | 4631 | |
d7ef4b73 | 4632 | do_ref: |
81714fb9 | 4633 | type = OP(scan); |
d7ef4b73 KW |
4634 | n = ARG(scan); /* which paren pair */ |
4635 | ||
4636 | do_nref_ref_common: | |
b93070ed | 4637 | ln = rex->offs[n].start; |
2c2d71f5 | 4638 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
b93070ed | 4639 | if (rex->lastparen < n || ln == -1) |
af3f8c16 | 4640 | sayNO; /* Do not match unless seen CLOSEn. */ |
b93070ed | 4641 | if (ln == rex->offs[n].end) |
a0d0e21e | 4642 | break; |
a0ed51b3 | 4643 | |
24d3c4a9 | 4644 | s = PL_bostr + ln; |
d7ef4b73 | 4645 | if (type != REF /* REF can do byte comparison */ |
2f65c56d | 4646 | && (utf8_target || type == REFFU)) |
d7ef4b73 KW |
4647 | { /* XXX handle REFFL better */ |
4648 | char * limit = PL_regeol; | |
4649 | ||
4650 | /* This call case insensitively compares the entire buffer | |
4651 | * at s, with the current input starting at locinput, but | |
4652 | * not going off the end given by PL_regeol, and returns in | |
2db97d41 | 4653 | * <limit> upon success, how much of the current input was |
d7ef4b73 | 4654 | * matched */ |
b93070ed | 4655 | if (! foldEQ_utf8_flags(s, NULL, rex->offs[n].end - ln, utf8_target, |
d513472c | 4656 | locinput, &limit, 0, utf8_target, utf8_fold_flags)) |
d7ef4b73 KW |
4657 | { |
4658 | sayNO; | |
a0ed51b3 | 4659 | } |
d7ef4b73 | 4660 | locinput = limit; |
a0ed51b3 LW |
4661 | break; |
4662 | } | |
4663 | ||
d7ef4b73 | 4664 | /* Not utf8: Inline the first character, for speed. */ |
7016d6eb DM |
4665 | if (!NEXTCHR_IS_EOS && |
4666 | UCHARAT(s) != nextchr && | |
81714fb9 | 4667 | (type == REF || |
d7ef4b73 | 4668 | UCHARAT(s) != fold_array[nextchr])) |
4633a7c4 | 4669 | sayNO; |
b93070ed | 4670 | ln = rex->offs[n].end - ln; |
24d3c4a9 | 4671 | if (locinput + ln > PL_regeol) |
4633a7c4 | 4672 | sayNO; |
81714fb9 | 4673 | if (ln > 1 && (type == REF |
24d3c4a9 | 4674 | ? memNE(s, locinput, ln) |
d7ef4b73 | 4675 | : ! folder(s, locinput, ln))) |
4633a7c4 | 4676 | sayNO; |
24d3c4a9 | 4677 | locinput += ln; |
a0d0e21e | 4678 | break; |
81714fb9 | 4679 | } |
3c0563b9 DM |
4680 | |
4681 | case NOTHING: /* null op; e.g. the 'nothing' following | |
4682 | * the '*' in m{(a+|b)*}' */ | |
4683 | break; | |
4684 | case TAIL: /* placeholder while compiling (A|B|C) */ | |
a0d0e21e | 4685 | break; |
3c0563b9 DM |
4686 | |
4687 | case BACK: /* ??? doesn't appear to be used ??? */ | |
a0d0e21e | 4688 | break; |
40a82448 DM |
4689 | |
4690 | #undef ST | |
4691 | #define ST st->u.eval | |
c277df42 | 4692 | { |
c277df42 | 4693 | SV *ret; |
d2f13c59 | 4694 | REGEXP *re_sv; |
6bda09f9 | 4695 | regexp *re; |
f8fc2ecf | 4696 | regexp_internal *rei; |
1a147d38 YO |
4697 | regnode *startpoint; |
4698 | ||
3c0563b9 | 4699 | case GOSTART: /* (?R) */ |
e7707071 YO |
4700 | case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */ |
4701 | if (cur_eval && cur_eval->locinput==locinput) { | |
24b23f37 | 4702 | if (cur_eval->u.eval.close_paren == (U32)ARG(scan)) |
1a147d38 | 4703 | Perl_croak(aTHX_ "Infinite recursion in regex"); |
4b196cd4 | 4704 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 YO |
4705 | Perl_croak(aTHX_ |
4706 | "Pattern subroutine nesting without pos change" | |
4707 | " exceeded limit in regex"); | |
6bda09f9 YO |
4708 | } else { |
4709 | nochange_depth = 0; | |
1a147d38 | 4710 | } |
288b8c02 | 4711 | re_sv = rex_sv; |
6bda09f9 | 4712 | re = rex; |
f8fc2ecf | 4713 | rei = rexi; |
1a147d38 | 4714 | if (OP(scan)==GOSUB) { |
6bda09f9 YO |
4715 | startpoint = scan + ARG2L(scan); |
4716 | ST.close_paren = ARG(scan); | |
4717 | } else { | |
f8fc2ecf | 4718 | startpoint = rei->program+1; |
6bda09f9 YO |
4719 | ST.close_paren = 0; |
4720 | } | |
4721 | goto eval_recurse_doit; | |
118e2215 | 4722 | assert(0); /* NOTREACHED */ |
3c0563b9 | 4723 | |
6bda09f9 YO |
4724 | case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */ |
4725 | if (cur_eval && cur_eval->locinput==locinput) { | |
4b196cd4 | 4726 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 | 4727 | Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex"); |
6bda09f9 YO |
4728 | } else { |
4729 | nochange_depth = 0; | |
4730 | } | |
8e5e9ebe | 4731 | { |
4aabdb9b | 4732 | /* execute the code in the {...} */ |
81ed78b2 | 4733 | |
4aabdb9b | 4734 | dSP; |
81ed78b2 | 4735 | SV ** before; |
1f4d1a1e | 4736 | OP * const oop = PL_op; |
4aabdb9b | 4737 | COP * const ocurcop = PL_curcop; |
81ed78b2 | 4738 | OP *nop; |
d80618d2 | 4739 | char *saved_regeol = PL_regeol; |
91332126 | 4740 | struct re_save_state saved_state; |
81ed78b2 | 4741 | CV *newcv; |
91332126 | 4742 | |
74088413 DM |
4743 | /* save *all* paren positions */ |
4744 | regcppush(rex, 0); | |
4745 | REGCP_SET(runops_cp); | |
4746 | ||
6562f1c4 | 4747 | /* To not corrupt the existing regex state while executing the |
b7f4cd04 FR |
4748 | * eval we would normally put it on the save stack, like with |
4749 | * save_re_context. However, re-evals have a weird scoping so we | |
4750 | * can't just add ENTER/LEAVE here. With that, things like | |
4751 | * | |
4752 | * (?{$a=2})(a(?{local$a=$a+1}))*aak*c(?{$b=$a}) | |
4753 | * | |
4754 | * would break, as they expect the localisation to be unwound | |
4755 | * only when the re-engine backtracks through the bit that | |
4756 | * localised it. | |
4757 | * | |
4758 | * What we do instead is just saving the state in a local c | |
4759 | * variable. | |
4760 | */ | |
91332126 | 4761 | Copy(&PL_reg_state, &saved_state, 1, struct re_save_state); |
81ed78b2 | 4762 | |
d24ca0c5 | 4763 | PL_reg_state.re_reparsing = FALSE; |
91332126 | 4764 | |
81ed78b2 DM |
4765 | if (!caller_cv) |
4766 | caller_cv = find_runcv(NULL); | |
4767 | ||
4aabdb9b | 4768 | n = ARG(scan); |
81ed78b2 | 4769 | |
b30fcab9 | 4770 | if (rexi->data->what[n] == 'r') { /* code from an external qr */ |
81ed78b2 | 4771 | newcv = ((struct regexp *)SvANY( |
b30fcab9 DM |
4772 | (REGEXP*)(rexi->data->data[n]) |
4773 | ))->qr_anoncv | |
81ed78b2 DM |
4774 | ; |
4775 | nop = (OP*)rexi->data->data[n+1]; | |
b30fcab9 DM |
4776 | } |
4777 | else if (rexi->data->what[n] == 'l') { /* literal code */ | |
81ed78b2 DM |
4778 | newcv = caller_cv; |
4779 | nop = (OP*)rexi->data->data[n]; | |
4780 | assert(CvDEPTH(newcv)); | |
68e2671b DM |
4781 | } |
4782 | else { | |
d24ca0c5 DM |
4783 | /* literal with own CV */ |
4784 | assert(rexi->data->what[n] == 'L'); | |
81ed78b2 DM |
4785 | newcv = rex->qr_anoncv; |
4786 | nop = (OP*)rexi->data->data[n]; | |
68e2671b | 4787 | } |
81ed78b2 | 4788 | |
0e458318 DM |
4789 | /* normally if we're about to execute code from the same |
4790 | * CV that we used previously, we just use the existing | |
4791 | * CX stack entry. However, its possible that in the | |
4792 | * meantime we may have backtracked, popped from the save | |
4793 | * stack, and undone the SAVECOMPPAD(s) associated with | |
4794 | * PUSH_MULTICALL; in which case PL_comppad no longer | |
4795 | * points to newcv's pad. */ | |
4796 | if (newcv != last_pushed_cv || PL_comppad != last_pad) | |
4797 | { | |
4798 | I32 depth = (newcv == caller_cv) ? 0 : 1; | |
4799 | if (last_pushed_cv) { | |
4800 | CHANGE_MULTICALL_WITHDEPTH(newcv, depth); | |
4801 | } | |
4802 | else { | |
4803 | PUSH_MULTICALL_WITHDEPTH(newcv, depth); | |
4804 | } | |
4805 | last_pushed_cv = newcv; | |
4806 | } | |
4807 | last_pad = PL_comppad; | |
4808 | ||
2e2e3f36 DM |
4809 | /* the initial nextstate you would normally execute |
4810 | * at the start of an eval (which would cause error | |
4811 | * messages to come from the eval), may be optimised | |
4812 | * away from the execution path in the regex code blocks; | |
4813 | * so manually set PL_curcop to it initially */ | |
4814 | { | |
81ed78b2 | 4815 | OP *o = cUNOPx(nop)->op_first; |
2e2e3f36 DM |
4816 | assert(o->op_type == OP_NULL); |
4817 | if (o->op_targ == OP_SCOPE) { | |
4818 | o = cUNOPo->op_first; | |
4819 | } | |
4820 | else { | |
4821 | assert(o->op_targ == OP_LEAVE); | |
4822 | o = cUNOPo->op_first; | |
4823 | assert(o->op_type == OP_ENTER); | |
4824 | o = o->op_sibling; | |
4825 | } | |
4826 | ||
4827 | if (o->op_type != OP_STUB) { | |
4828 | assert( o->op_type == OP_NEXTSTATE | |
4829 | || o->op_type == OP_DBSTATE | |
4830 | || (o->op_type == OP_NULL | |
4831 | && ( o->op_targ == OP_NEXTSTATE | |
4832 | || o->op_targ == OP_DBSTATE | |
4833 | ) | |
4834 | ) | |
4835 | ); | |
4836 | PL_curcop = (COP*)o; | |
4837 | } | |
4838 | } | |
81ed78b2 | 4839 | nop = nop->op_next; |
2e2e3f36 | 4840 | |
24b23f37 | 4841 | DEBUG_STATE_r( PerlIO_printf(Perl_debug_log, |
81ed78b2 DM |
4842 | " re EVAL PL_op=0x%"UVxf"\n", PTR2UV(nop)) ); |
4843 | ||
b93070ed | 4844 | rex->offs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr; |
4aabdb9b | 4845 | |
2bf803e2 YO |
4846 | if (sv_yes_mark) { |
4847 | SV *sv_mrk = get_sv("REGMARK", 1); | |
4848 | sv_setsv(sv_mrk, sv_yes_mark); | |
4849 | } | |
4850 | ||
81ed78b2 DM |
4851 | /* we don't use MULTICALL here as we want to call the |
4852 | * first op of the block of interest, rather than the | |
4853 | * first op of the sub */ | |
4854 | before = SP; | |
4855 | PL_op = nop; | |
8e5e9ebe RGS |
4856 | CALLRUNOPS(aTHX); /* Scalar context. */ |
4857 | SPAGAIN; | |
4858 | if (SP == before) | |
075aa684 | 4859 | ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */ |
8e5e9ebe RGS |
4860 | else { |
4861 | ret = POPs; | |
4862 | PUTBACK; | |
4863 | } | |
4aabdb9b | 4864 | |
e4bfbed3 DM |
4865 | /* before restoring everything, evaluate the returned |
4866 | * value, so that 'uninit' warnings don't use the wrong | |
497d0a96 DM |
4867 | * PL_op or pad. Also need to process any magic vars |
4868 | * (e.g. $1) *before* parentheses are restored */ | |
e4bfbed3 DM |
4869 | |
4870 | PL_op = NULL; | |
4871 | ||
5e98dac2 | 4872 | re_sv = NULL; |
e4bfbed3 DM |
4873 | if (logical == 0) /* (?{})/ */ |
4874 | sv_setsv(save_scalar(PL_replgv), ret); /* $^R */ | |
4875 | else if (logical == 1) { /* /(?(?{...})X|Y)/ */ | |
4876 | sw = cBOOL(SvTRUE(ret)); | |
4877 | logical = 0; | |
4878 | } | |
4879 | else { /* /(??{}) */ | |
497d0a96 DM |
4880 | /* if its overloaded, let the regex compiler handle |
4881 | * it; otherwise extract regex, or stringify */ | |
4882 | if (!SvAMAGIC(ret)) { | |
4883 | SV *sv = ret; | |
4884 | if (SvROK(sv)) | |
4885 | sv = SvRV(sv); | |
4886 | if (SvTYPE(sv) == SVt_REGEXP) | |
4887 | re_sv = (REGEXP*) sv; | |
4888 | else if (SvSMAGICAL(sv)) { | |
4889 | MAGIC *mg = mg_find(sv, PERL_MAGIC_qr); | |
4890 | if (mg) | |
4891 | re_sv = (REGEXP *) mg->mg_obj; | |
4892 | } | |
e4bfbed3 | 4893 | |
497d0a96 DM |
4894 | /* force any magic, undef warnings here */ |
4895 | if (!re_sv) { | |
4896 | ret = sv_mortalcopy(ret); | |
4897 | (void) SvPV_force_nolen(ret); | |
4898 | } | |
e4bfbed3 DM |
4899 | } |
4900 | ||
4901 | } | |
4902 | ||
91332126 FR |
4903 | Copy(&saved_state, &PL_reg_state, 1, struct re_save_state); |
4904 | ||
81ed78b2 DM |
4905 | /* *** Note that at this point we don't restore |
4906 | * PL_comppad, (or pop the CxSUB) on the assumption it may | |
4907 | * be used again soon. This is safe as long as nothing | |
4908 | * in the regexp code uses the pad ! */ | |
4aabdb9b | 4909 | PL_op = oop; |
4aabdb9b | 4910 | PL_curcop = ocurcop; |
d80618d2 | 4911 | PL_regeol = saved_regeol; |
e4bfbed3 DM |
4912 | S_regcp_restore(aTHX_ rex, runops_cp); |
4913 | ||
4914 | if (logical != 2) | |
4aabdb9b | 4915 | break; |
8e5e9ebe | 4916 | } |
e4bfbed3 DM |
4917 | |
4918 | /* only /(??{})/ from now on */ | |
24d3c4a9 | 4919 | logical = 0; |
4aabdb9b | 4920 | { |
4f639d21 DM |
4921 | /* extract RE object from returned value; compiling if |
4922 | * necessary */ | |
5c35adbb | 4923 | |
575c37f6 DM |
4924 | if (re_sv) { |
4925 | re_sv = reg_temp_copy(NULL, re_sv); | |
288b8c02 | 4926 | } |
0f5d15d6 | 4927 | else { |
c737faaf | 4928 | U32 pm_flags = 0; |
a3b680e6 | 4929 | const I32 osize = PL_regsize; |
0f5d15d6 | 4930 | |
9753d940 DM |
4931 | if (SvUTF8(ret) && IN_BYTES) { |
4932 | /* In use 'bytes': make a copy of the octet | |
4933 | * sequence, but without the flag on */ | |
b9ad30b4 NC |
4934 | STRLEN len; |
4935 | const char *const p = SvPV(ret, len); | |
4936 | ret = newSVpvn_flags(p, len, SVs_TEMP); | |
4937 | } | |
732caac7 DM |
4938 | if (rex->intflags & PREGf_USE_RE_EVAL) |
4939 | pm_flags |= PMf_USE_RE_EVAL; | |
4940 | ||
4941 | /* if we got here, it should be an engine which | |
4942 | * supports compiling code blocks and stuff */ | |
4943 | assert(rex->engine && rex->engine->op_comp); | |
ec841a27 | 4944 | assert(!(scan->flags & ~RXf_PMf_COMPILETIME)); |
575c37f6 | 4945 | re_sv = rex->engine->op_comp(aTHX_ &ret, 1, NULL, |
ec841a27 DM |
4946 | rex->engine, NULL, NULL, |
4947 | /* copy /msix etc to inner pattern */ | |
4948 | scan->flags, | |
4949 | pm_flags); | |
732caac7 | 4950 | |
9041c2e3 | 4951 | if (!(SvFLAGS(ret) |
faf82a0b | 4952 | & (SVs_TEMP | SVs_PADTMP | SVf_READONLY |
3ce3ed55 | 4953 | | SVs_GMG))) { |
a2794585 NC |
4954 | /* This isn't a first class regexp. Instead, it's |
4955 | caching a regexp onto an existing, Perl visible | |
4956 | scalar. */ | |
575c37f6 | 4957 | sv_magic(ret, MUTABLE_SV(re_sv), PERL_MAGIC_qr, 0, 0); |
3ce3ed55 | 4958 | } |
0f5d15d6 | 4959 | PL_regsize = osize; |
74088413 DM |
4960 | /* safe to do now that any $1 etc has been |
4961 | * interpolated into the new pattern string and | |
4962 | * compiled */ | |
4963 | S_regcp_restore(aTHX_ rex, runops_cp); | |
0f5d15d6 | 4964 | } |
575c37f6 | 4965 | re = (struct regexp *)SvANY(re_sv); |
4aabdb9b | 4966 | } |
07bc277f | 4967 | RXp_MATCH_COPIED_off(re); |
28d8d7f4 YO |
4968 | re->subbeg = rex->subbeg; |
4969 | re->sublen = rex->sublen; | |
6502e081 DM |
4970 | re->suboffset = rex->suboffset; |
4971 | re->subcoffset = rex->subcoffset; | |
f8fc2ecf | 4972 | rei = RXi_GET(re); |
6bda09f9 | 4973 | DEBUG_EXECUTE_r( |
f2ed9b32 | 4974 | debug_start_match(re_sv, utf8_target, locinput, PL_regeol, |
6bda09f9 YO |
4975 | "Matching embedded"); |
4976 | ); | |
f8fc2ecf | 4977 | startpoint = rei->program + 1; |
1a147d38 | 4978 | ST.close_paren = 0; /* only used for GOSUB */ |
aa283a38 | 4979 | |
1a147d38 | 4980 | eval_recurse_doit: /* Share code with GOSUB below this line */ |
aa283a38 | 4981 | /* run the pattern returned from (??{...}) */ |
b93070ed | 4982 | ST.cp = regcppush(rex, 0); /* Save *all* the positions. */ |
40a82448 | 4983 | REGCP_SET(ST.lastcp); |
6bda09f9 | 4984 | |
0357f1fd ML |
4985 | re->lastparen = 0; |
4986 | re->lastcloseparen = 0; | |
4987 | ||
ae0beba1 | 4988 | PL_regsize = 0; |
4aabdb9b DM |
4989 | |
4990 | /* XXXX This is too dramatic a measure... */ | |
4991 | PL_reg_maxiter = 0; | |
4992 | ||
faec1544 | 4993 | ST.toggle_reg_flags = PL_reg_flags; |
3c8556c3 | 4994 | if (RX_UTF8(re_sv)) |
faec1544 DM |
4995 | PL_reg_flags |= RF_utf8; |
4996 | else | |
4997 | PL_reg_flags &= ~RF_utf8; | |
4998 | ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */ | |
4999 | ||
288b8c02 | 5000 | ST.prev_rex = rex_sv; |
faec1544 | 5001 | ST.prev_curlyx = cur_curlyx; |
ec43f78b DM |
5002 | rex_sv = re_sv; |
5003 | SET_reg_curpm(rex_sv); | |
288b8c02 | 5004 | rex = re; |
f8fc2ecf | 5005 | rexi = rei; |
faec1544 | 5006 | cur_curlyx = NULL; |
40a82448 | 5007 | ST.B = next; |
faec1544 DM |
5008 | ST.prev_eval = cur_eval; |
5009 | cur_eval = st; | |
faec1544 | 5010 | /* now continue from first node in postoned RE */ |
4d5016e5 | 5011 | PUSH_YES_STATE_GOTO(EVAL_AB, startpoint, locinput); |
118e2215 | 5012 | assert(0); /* NOTREACHED */ |
c277df42 | 5013 | } |
40a82448 | 5014 | |
faec1544 DM |
5015 | case EVAL_AB: /* cleanup after a successful (??{A})B */ |
5016 | /* note: this is called twice; first after popping B, then A */ | |
5017 | PL_reg_flags ^= ST.toggle_reg_flags; | |
ec43f78b DM |
5018 | rex_sv = ST.prev_rex; |
5019 | SET_reg_curpm(rex_sv); | |
288b8c02 | 5020 | rex = (struct regexp *)SvANY(rex_sv); |
f8fc2ecf | 5021 | rexi = RXi_GET(rex); |
faec1544 DM |
5022 | regcpblow(ST.cp); |
5023 | cur_eval = ST.prev_eval; | |
5024 | cur_curlyx = ST.prev_curlyx; | |
34a81e2b | 5025 | |
40a82448 DM |
5026 | /* XXXX This is too dramatic a measure... */ |
5027 | PL_reg_maxiter = 0; | |
e7707071 | 5028 | if ( nochange_depth ) |
4b196cd4 | 5029 | nochange_depth--; |
262b90c4 | 5030 | sayYES; |
40a82448 | 5031 | |
40a82448 | 5032 | |
faec1544 DM |
5033 | case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */ |
5034 | /* note: this is called twice; first after popping B, then A */ | |
5035 | PL_reg_flags ^= ST.toggle_reg_flags; | |
ec43f78b DM |
5036 | rex_sv = ST.prev_rex; |
5037 | SET_reg_curpm(rex_sv); | |
288b8c02 | 5038 | rex = (struct regexp *)SvANY(rex_sv); |
f8fc2ecf | 5039 | rexi = RXi_GET(rex); |
0357f1fd | 5040 | |
40a82448 DM |
5041 | REGCP_UNWIND(ST.lastcp); |
5042 | regcppop(rex); | |
faec1544 DM |
5043 | cur_eval = ST.prev_eval; |
5044 | cur_curlyx = ST.prev_curlyx; | |
5045 | /* XXXX This is too dramatic a measure... */ | |
5046 | PL_reg_maxiter = 0; | |
e7707071 | 5047 | if ( nochange_depth ) |
4b196cd4 | 5048 | nochange_depth--; |
40a82448 | 5049 | sayNO_SILENT; |
40a82448 DM |
5050 | #undef ST |
5051 | ||
3c0563b9 | 5052 | case OPEN: /* ( */ |
c277df42 | 5053 | n = ARG(scan); /* which paren pair */ |
1ca2007e | 5054 | rex->offs[n].start_tmp = locinput - PL_bostr; |
3280af22 NIS |
5055 | if (n > PL_regsize) |
5056 | PL_regsize = n; | |
495f47a5 DM |
5057 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
5058 | "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf" tmp; regsize=%"UVuf"\n", | |
5059 | PTR2UV(rex), | |
5060 | PTR2UV(rex->offs), | |
5061 | (UV)n, | |
5062 | (IV)rex->offs[n].start_tmp, | |
5063 | (UV)PL_regsize | |
5064 | )); | |
e2e6a0f1 | 5065 | lastopen = n; |
a0d0e21e | 5066 | break; |
495f47a5 DM |
5067 | |
5068 | /* XXX really need to log other places start/end are set too */ | |
5069 | #define CLOSE_CAPTURE \ | |
5070 | rex->offs[n].start = rex->offs[n].start_tmp; \ | |
5071 | rex->offs[n].end = locinput - PL_bostr; \ | |
5072 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, \ | |
5073 | "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf"..%"IVdf"\n", \ | |
5074 | PTR2UV(rex), \ | |
5075 | PTR2UV(rex->offs), \ | |
5076 | (UV)n, \ | |
5077 | (IV)rex->offs[n].start, \ | |
5078 | (IV)rex->offs[n].end \ | |
5079 | )) | |
5080 | ||
3c0563b9 | 5081 | case CLOSE: /* ) */ |
c277df42 | 5082 | n = ARG(scan); /* which paren pair */ |
495f47a5 | 5083 | CLOSE_CAPTURE; |
7f69552c YO |
5084 | /*if (n > PL_regsize) |
5085 | PL_regsize = n;*/ | |
b93070ed DM |
5086 | if (n > rex->lastparen) |
5087 | rex->lastparen = n; | |
5088 | rex->lastcloseparen = n; | |
3b6647e0 | 5089 | if (cur_eval && cur_eval->u.eval.close_paren == n) { |
6bda09f9 YO |
5090 | goto fake_end; |
5091 | } | |
a0d0e21e | 5092 | break; |
3c0563b9 DM |
5093 | |
5094 | case ACCEPT: /* (*ACCEPT) */ | |
e2e6a0f1 YO |
5095 | if (ARG(scan)){ |
5096 | regnode *cursor; | |
5097 | for (cursor=scan; | |
5098 | cursor && OP(cursor)!=END; | |
5099 | cursor=regnext(cursor)) | |
5100 | { | |
5101 | if ( OP(cursor)==CLOSE ){ | |
5102 | n = ARG(cursor); | |
5103 | if ( n <= lastopen ) { | |
495f47a5 | 5104 | CLOSE_CAPTURE; |
e2e6a0f1 YO |
5105 | /*if (n > PL_regsize) |
5106 | PL_regsize = n;*/ | |
b93070ed DM |
5107 | if (n > rex->lastparen) |
5108 | rex->lastparen = n; | |
5109 | rex->lastcloseparen = n; | |
3b6647e0 RB |
5110 | if ( n == ARG(scan) || (cur_eval && |
5111 | cur_eval->u.eval.close_paren == n)) | |
e2e6a0f1 YO |
5112 | break; |
5113 | } | |
5114 | } | |
5115 | } | |
5116 | } | |
5117 | goto fake_end; | |
5118 | /*NOTREACHED*/ | |
3c0563b9 DM |
5119 | |
5120 | case GROUPP: /* (?(1)) */ | |
c277df42 | 5121 | n = ARG(scan); /* which paren pair */ |
b93070ed | 5122 | sw = cBOOL(rex->lastparen >= n && rex->offs[n].end != -1); |
c277df42 | 5123 | break; |
3c0563b9 DM |
5124 | |
5125 | case NGROUPP: /* (?(<name>)) */ | |
0a4db386 | 5126 | /* reg_check_named_buff_matched returns 0 for no match */ |
f2338a2e | 5127 | sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan)); |
0a4db386 | 5128 | break; |
3c0563b9 DM |
5129 | |
5130 | case INSUBP: /* (?(R)) */ | |
0a4db386 | 5131 | n = ARG(scan); |
3b6647e0 | 5132 | sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n)); |
0a4db386 | 5133 | break; |
3c0563b9 DM |
5134 | |
5135 | case DEFINEP: /* (?(DEFINE)) */ | |
0a4db386 YO |
5136 | sw = 0; |
5137 | break; | |
3c0563b9 DM |
5138 | |
5139 | case IFTHEN: /* (?(cond)A|B) */ | |
2c2d71f5 | 5140 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
24d3c4a9 | 5141 | if (sw) |
c277df42 IZ |
5142 | next = NEXTOPER(NEXTOPER(scan)); |
5143 | else { | |
5144 | next = scan + ARG(scan); | |
5145 | if (OP(next) == IFTHEN) /* Fake one. */ | |
5146 | next = NEXTOPER(NEXTOPER(next)); | |
5147 | } | |
5148 | break; | |
3c0563b9 DM |
5149 | |
5150 | case LOGICAL: /* modifier for EVAL and IFMATCH */ | |
24d3c4a9 | 5151 | logical = scan->flags; |
c277df42 | 5152 | break; |
c476f425 | 5153 | |
2ab05381 | 5154 | /******************************************************************* |
2ab05381 | 5155 | |
c476f425 DM |
5156 | The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/ |
5157 | pattern, where A and B are subpatterns. (For simple A, CURLYM or | |
5158 | STAR/PLUS/CURLY/CURLYN are used instead.) | |
2ab05381 | 5159 | |
c476f425 | 5160 | A*B is compiled as <CURLYX><A><WHILEM><B> |
2ab05381 | 5161 | |
c476f425 DM |
5162 | On entry to the subpattern, CURLYX is called. This pushes a CURLYX |
5163 | state, which contains the current count, initialised to -1. It also sets | |
5164 | cur_curlyx to point to this state, with any previous value saved in the | |
5165 | state block. | |
2ab05381 | 5166 | |
c476f425 DM |
5167 | CURLYX then jumps straight to the WHILEM op, rather than executing A, |
5168 | since the pattern may possibly match zero times (i.e. it's a while {} loop | |
5169 | rather than a do {} while loop). | |
2ab05381 | 5170 | |
c476f425 DM |
5171 | Each entry to WHILEM represents a successful match of A. The count in the |
5172 | CURLYX block is incremented, another WHILEM state is pushed, and execution | |
5173 | passes to A or B depending on greediness and the current count. | |
2ab05381 | 5174 | |
c476f425 DM |
5175 | For example, if matching against the string a1a2a3b (where the aN are |
5176 | substrings that match /A/), then the match progresses as follows: (the | |
5177 | pushed states are interspersed with the bits of strings matched so far): | |
2ab05381 | 5178 | |
c476f425 DM |
5179 | <CURLYX cnt=-1> |
5180 | <CURLYX cnt=0><WHILEM> | |
5181 | <CURLYX cnt=1><WHILEM> a1 <WHILEM> | |
5182 | <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM> | |
5183 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> | |
5184 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b | |
2ab05381 | 5185 | |
c476f425 DM |
5186 | (Contrast this with something like CURLYM, which maintains only a single |
5187 | backtrack state: | |
2ab05381 | 5188 | |
c476f425 DM |
5189 | <CURLYM cnt=0> a1 |
5190 | a1 <CURLYM cnt=1> a2 | |
5191 | a1 a2 <CURLYM cnt=2> a3 | |
5192 | a1 a2 a3 <CURLYM cnt=3> b | |
5193 | ) | |
2ab05381 | 5194 | |
c476f425 DM |
5195 | Each WHILEM state block marks a point to backtrack to upon partial failure |
5196 | of A or B, and also contains some minor state data related to that | |
5197 | iteration. The CURLYX block, pointed to by cur_curlyx, contains the | |
5198 | overall state, such as the count, and pointers to the A and B ops. | |
2ab05381 | 5199 | |
c476f425 DM |
5200 | This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx |
5201 | must always point to the *current* CURLYX block, the rules are: | |
2ab05381 | 5202 | |
c476f425 DM |
5203 | When executing CURLYX, save the old cur_curlyx in the CURLYX state block, |
5204 | and set cur_curlyx to point the new block. | |
2ab05381 | 5205 | |
c476f425 DM |
5206 | When popping the CURLYX block after a successful or unsuccessful match, |
5207 | restore the previous cur_curlyx. | |
2ab05381 | 5208 | |
c476f425 DM |
5209 | When WHILEM is about to execute B, save the current cur_curlyx, and set it |
5210 | to the outer one saved in the CURLYX block. | |
2ab05381 | 5211 | |
c476f425 DM |
5212 | When popping the WHILEM block after a successful or unsuccessful B match, |
5213 | restore the previous cur_curlyx. | |
2ab05381 | 5214 | |
c476f425 DM |
5215 | Here's an example for the pattern (AI* BI)*BO |
5216 | I and O refer to inner and outer, C and W refer to CURLYX and WHILEM: | |
2ab05381 | 5217 | |
c476f425 DM |
5218 | cur_ |
5219 | curlyx backtrack stack | |
5220 | ------ --------------- | |
5221 | NULL | |
5222 | CO <CO prev=NULL> <WO> | |
5223 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
5224 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi | |
5225 | NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo | |
2ab05381 | 5226 | |
c476f425 DM |
5227 | At this point the pattern succeeds, and we work back down the stack to |
5228 | clean up, restoring as we go: | |
95b24440 | 5229 | |
c476f425 DM |
5230 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi |
5231 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
5232 | CO <CO prev=NULL> <WO> | |
5233 | NULL | |
a0374537 | 5234 | |
c476f425 DM |
5235 | *******************************************************************/ |
5236 | ||
5237 | #define ST st->u.curlyx | |
5238 | ||
5239 | case CURLYX: /* start of /A*B/ (for complex A) */ | |
5240 | { | |
5241 | /* No need to save/restore up to this paren */ | |
5242 | I32 parenfloor = scan->flags; | |
5243 | ||
5244 | assert(next); /* keep Coverity happy */ | |
5245 | if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ | |
5246 | next += ARG(next); | |
5247 | ||
5248 | /* XXXX Probably it is better to teach regpush to support | |
5249 | parenfloor > PL_regsize... */ | |
b93070ed DM |
5250 | if (parenfloor > (I32)rex->lastparen) |
5251 | parenfloor = rex->lastparen; /* Pessimization... */ | |
c476f425 DM |
5252 | |
5253 | ST.prev_curlyx= cur_curlyx; | |
5254 | cur_curlyx = st; | |
5255 | ST.cp = PL_savestack_ix; | |
5256 | ||
5257 | /* these fields contain the state of the current curly. | |
5258 | * they are accessed by subsequent WHILEMs */ | |
5259 | ST.parenfloor = parenfloor; | |
d02d6d97 | 5260 | ST.me = scan; |
c476f425 | 5261 | ST.B = next; |
24d3c4a9 DM |
5262 | ST.minmod = minmod; |
5263 | minmod = 0; | |
c476f425 DM |
5264 | ST.count = -1; /* this will be updated by WHILEM */ |
5265 | ST.lastloc = NULL; /* this will be updated by WHILEM */ | |
5266 | ||
4d5016e5 | 5267 | PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next), locinput); |
118e2215 | 5268 | assert(0); /* NOTREACHED */ |
c476f425 | 5269 | } |
a0d0e21e | 5270 | |
c476f425 | 5271 | case CURLYX_end: /* just finished matching all of A*B */ |
c476f425 DM |
5272 | cur_curlyx = ST.prev_curlyx; |
5273 | sayYES; | |
118e2215 | 5274 | assert(0); /* NOTREACHED */ |
a0d0e21e | 5275 | |
c476f425 DM |
5276 | case CURLYX_end_fail: /* just failed to match all of A*B */ |
5277 | regcpblow(ST.cp); | |
5278 | cur_curlyx = ST.prev_curlyx; | |
5279 | sayNO; | |
118e2215 | 5280 | assert(0); /* NOTREACHED */ |
4633a7c4 | 5281 | |
a0d0e21e | 5282 | |
c476f425 DM |
5283 | #undef ST |
5284 | #define ST st->u.whilem | |
5285 | ||
5286 | case WHILEM: /* just matched an A in /A*B/ (for complex A) */ | |
5287 | { | |
5288 | /* see the discussion above about CURLYX/WHILEM */ | |
c476f425 | 5289 | I32 n; |
d02d6d97 DM |
5290 | int min = ARG1(cur_curlyx->u.curlyx.me); |
5291 | int max = ARG2(cur_curlyx->u.curlyx.me); | |
5292 | regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS; | |
5293 | ||
c476f425 DM |
5294 | assert(cur_curlyx); /* keep Coverity happy */ |
5295 | n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */ | |
5296 | ST.save_lastloc = cur_curlyx->u.curlyx.lastloc; | |
5297 | ST.cache_offset = 0; | |
5298 | ST.cache_mask = 0; | |
5299 | ||
c476f425 DM |
5300 | |
5301 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
d02d6d97 DM |
5302 | "%*s whilem: matched %ld out of %d..%d\n", |
5303 | REPORT_CODE_OFF+depth*2, "", (long)n, min, max) | |
c476f425 | 5304 | ); |
a0d0e21e | 5305 | |
c476f425 | 5306 | /* First just match a string of min A's. */ |
a0d0e21e | 5307 | |
d02d6d97 | 5308 | if (n < min) { |
b93070ed | 5309 | ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor); |
c476f425 | 5310 | cur_curlyx->u.curlyx.lastloc = locinput; |
92e82afa YO |
5311 | REGCP_SET(ST.lastcp); |
5312 | ||
4d5016e5 | 5313 | PUSH_STATE_GOTO(WHILEM_A_pre, A, locinput); |
118e2215 | 5314 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5315 | } |
5316 | ||
5317 | /* If degenerate A matches "", assume A done. */ | |
5318 | ||
5319 | if (locinput == cur_curlyx->u.curlyx.lastloc) { | |
5320 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
5321 | "%*s whilem: empty match detected, trying continuation...\n", | |
5322 | REPORT_CODE_OFF+depth*2, "") | |
5323 | ); | |
5324 | goto do_whilem_B_max; | |
5325 | } | |
5326 | ||
5327 | /* super-linear cache processing */ | |
5328 | ||
5329 | if (scan->flags) { | |
a0d0e21e | 5330 | |
2c2d71f5 | 5331 | if (!PL_reg_maxiter) { |
c476f425 DM |
5332 | /* start the countdown: Postpone detection until we |
5333 | * know the match is not *that* much linear. */ | |
2c2d71f5 | 5334 | PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4); |
66bf836d DM |
5335 | /* possible overflow for long strings and many CURLYX's */ |
5336 | if (PL_reg_maxiter < 0) | |
5337 | PL_reg_maxiter = I32_MAX; | |
2c2d71f5 JH |
5338 | PL_reg_leftiter = PL_reg_maxiter; |
5339 | } | |
c476f425 | 5340 | |
2c2d71f5 | 5341 | if (PL_reg_leftiter-- == 0) { |
c476f425 | 5342 | /* initialise cache */ |
3298f257 | 5343 | const I32 size = (PL_reg_maxiter + 7)/8; |
2c2d71f5 | 5344 | if (PL_reg_poscache) { |
eb160463 | 5345 | if ((I32)PL_reg_poscache_size < size) { |
2c2d71f5 JH |
5346 | Renew(PL_reg_poscache, size, char); |
5347 | PL_reg_poscache_size = size; | |
5348 | } | |
5349 | Zero(PL_reg_poscache, size, char); | |
5350 | } | |
5351 | else { | |
5352 | PL_reg_poscache_size = size; | |
a02a5408 | 5353 | Newxz(PL_reg_poscache, size, char); |
2c2d71f5 | 5354 | } |
c476f425 DM |
5355 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
5356 | "%swhilem: Detected a super-linear match, switching on caching%s...\n", | |
5357 | PL_colors[4], PL_colors[5]) | |
5358 | ); | |
2c2d71f5 | 5359 | } |
c476f425 | 5360 | |
2c2d71f5 | 5361 | if (PL_reg_leftiter < 0) { |
c476f425 DM |
5362 | /* have we already failed at this position? */ |
5363 | I32 offset, mask; | |
5364 | offset = (scan->flags & 0xf) - 1 | |
5365 | + (locinput - PL_bostr) * (scan->flags>>4); | |
5366 | mask = 1 << (offset % 8); | |
5367 | offset /= 8; | |
5368 | if (PL_reg_poscache[offset] & mask) { | |
5369 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
5370 | "%*s whilem: (cache) already tried at this position...\n", | |
5371 | REPORT_CODE_OFF+depth*2, "") | |
2c2d71f5 | 5372 | ); |
3298f257 | 5373 | sayNO; /* cache records failure */ |
2c2d71f5 | 5374 | } |
c476f425 DM |
5375 | ST.cache_offset = offset; |
5376 | ST.cache_mask = mask; | |
2c2d71f5 | 5377 | } |
c476f425 | 5378 | } |
2c2d71f5 | 5379 | |
c476f425 | 5380 | /* Prefer B over A for minimal matching. */ |
a687059c | 5381 | |
c476f425 DM |
5382 | if (cur_curlyx->u.curlyx.minmod) { |
5383 | ST.save_curlyx = cur_curlyx; | |
5384 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
b93070ed | 5385 | ST.cp = regcppush(rex, ST.save_curlyx->u.curlyx.parenfloor); |
c476f425 | 5386 | REGCP_SET(ST.lastcp); |
4d5016e5 DM |
5387 | PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B, |
5388 | locinput); | |
118e2215 | 5389 | assert(0); /* NOTREACHED */ |
c476f425 | 5390 | } |
a0d0e21e | 5391 | |
c476f425 DM |
5392 | /* Prefer A over B for maximal matching. */ |
5393 | ||
d02d6d97 | 5394 | if (n < max) { /* More greed allowed? */ |
b93070ed | 5395 | ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor); |
c476f425 DM |
5396 | cur_curlyx->u.curlyx.lastloc = locinput; |
5397 | REGCP_SET(ST.lastcp); | |
4d5016e5 | 5398 | PUSH_STATE_GOTO(WHILEM_A_max, A, locinput); |
118e2215 | 5399 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5400 | } |
5401 | goto do_whilem_B_max; | |
5402 | } | |
118e2215 | 5403 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5404 | |
5405 | case WHILEM_B_min: /* just matched B in a minimal match */ | |
5406 | case WHILEM_B_max: /* just matched B in a maximal match */ | |
5407 | cur_curlyx = ST.save_curlyx; | |
5408 | sayYES; | |
118e2215 | 5409 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5410 | |
5411 | case WHILEM_B_max_fail: /* just failed to match B in a maximal match */ | |
5412 | cur_curlyx = ST.save_curlyx; | |
5413 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; | |
5414 | cur_curlyx->u.curlyx.count--; | |
5415 | CACHEsayNO; | |
118e2215 | 5416 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5417 | |
5418 | case WHILEM_A_min_fail: /* just failed to match A in a minimal match */ | |
c476f425 DM |
5419 | /* FALL THROUGH */ |
5420 | case WHILEM_A_pre_fail: /* just failed to match even minimal A */ | |
92e82afa YO |
5421 | REGCP_UNWIND(ST.lastcp); |
5422 | regcppop(rex); | |
c476f425 DM |
5423 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; |
5424 | cur_curlyx->u.curlyx.count--; | |
5425 | CACHEsayNO; | |
118e2215 | 5426 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5427 | |
5428 | case WHILEM_A_max_fail: /* just failed to match A in a maximal match */ | |
5429 | REGCP_UNWIND(ST.lastcp); | |
5430 | regcppop(rex); /* Restore some previous $<digit>s? */ | |
c476f425 DM |
5431 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
5432 | "%*s whilem: failed, trying continuation...\n", | |
5433 | REPORT_CODE_OFF+depth*2, "") | |
5434 | ); | |
5435 | do_whilem_B_max: | |
5436 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
5437 | && ckWARN(WARN_REGEXP) | |
5438 | && !(PL_reg_flags & RF_warned)) | |
5439 | { | |
5440 | PL_reg_flags |= RF_warned; | |
dcbac5bb FC |
5441 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), |
5442 | "Complex regular subexpression recursion limit (%d) " | |
5443 | "exceeded", | |
c476f425 DM |
5444 | REG_INFTY - 1); |
5445 | } | |
5446 | ||
5447 | /* now try B */ | |
5448 | ST.save_curlyx = cur_curlyx; | |
5449 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4d5016e5 DM |
5450 | PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B, |
5451 | locinput); | |
118e2215 | 5452 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5453 | |
5454 | case WHILEM_B_min_fail: /* just failed to match B in a minimal match */ | |
5455 | cur_curlyx = ST.save_curlyx; | |
5456 | REGCP_UNWIND(ST.lastcp); | |
5457 | regcppop(rex); | |
5458 | ||
d02d6d97 | 5459 | if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) { |
c476f425 DM |
5460 | /* Maximum greed exceeded */ |
5461 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
5462 | && ckWARN(WARN_REGEXP) | |
5463 | && !(PL_reg_flags & RF_warned)) | |
5464 | { | |
3280af22 | 5465 | PL_reg_flags |= RF_warned; |
c476f425 | 5466 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), |
dcbac5bb FC |
5467 | "Complex regular subexpression recursion " |
5468 | "limit (%d) exceeded", | |
c476f425 | 5469 | REG_INFTY - 1); |
a0d0e21e | 5470 | } |
c476f425 | 5471 | cur_curlyx->u.curlyx.count--; |
3ab3c9b4 | 5472 | CACHEsayNO; |
a0d0e21e | 5473 | } |
c476f425 DM |
5474 | |
5475 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
5476 | "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "") | |
5477 | ); | |
5478 | /* Try grabbing another A and see if it helps. */ | |
c476f425 | 5479 | cur_curlyx->u.curlyx.lastloc = locinput; |
b93070ed | 5480 | ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor); |
c476f425 | 5481 | REGCP_SET(ST.lastcp); |
d02d6d97 | 5482 | PUSH_STATE_GOTO(WHILEM_A_min, |
4d5016e5 DM |
5483 | /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS, |
5484 | locinput); | |
118e2215 | 5485 | assert(0); /* NOTREACHED */ |
40a82448 DM |
5486 | |
5487 | #undef ST | |
5488 | #define ST st->u.branch | |
5489 | ||
5490 | case BRANCHJ: /* /(...|A|...)/ with long next pointer */ | |
c277df42 IZ |
5491 | next = scan + ARG(scan); |
5492 | if (next == scan) | |
5493 | next = NULL; | |
40a82448 DM |
5494 | scan = NEXTOPER(scan); |
5495 | /* FALL THROUGH */ | |
c277df42 | 5496 | |
40a82448 DM |
5497 | case BRANCH: /* /(...|A|...)/ */ |
5498 | scan = NEXTOPER(scan); /* scan now points to inner node */ | |
b93070ed | 5499 | ST.lastparen = rex->lastparen; |
f6033a9d | 5500 | ST.lastcloseparen = rex->lastcloseparen; |
40a82448 DM |
5501 | ST.next_branch = next; |
5502 | REGCP_SET(ST.cp); | |
02db2b7b | 5503 | |
40a82448 | 5504 | /* Now go into the branch */ |
5d458dd8 | 5505 | if (has_cutgroup) { |
4d5016e5 | 5506 | PUSH_YES_STATE_GOTO(BRANCH_next, scan, locinput); |
5d458dd8 | 5507 | } else { |
4d5016e5 | 5508 | PUSH_STATE_GOTO(BRANCH_next, scan, locinput); |
5d458dd8 | 5509 | } |
118e2215 | 5510 | assert(0); /* NOTREACHED */ |
3c0563b9 DM |
5511 | |
5512 | case CUTGROUP: /* /(*THEN)/ */ | |
5d458dd8 | 5513 | sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL : |
ad64d0ec | 5514 | MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
4d5016e5 | 5515 | PUSH_STATE_GOTO(CUTGROUP_next, next, locinput); |
118e2215 | 5516 | assert(0); /* NOTREACHED */ |
3c0563b9 | 5517 | |
5d458dd8 YO |
5518 | case CUTGROUP_next_fail: |
5519 | do_cutgroup = 1; | |
5520 | no_final = 1; | |
5521 | if (st->u.mark.mark_name) | |
5522 | sv_commit = st->u.mark.mark_name; | |
5523 | sayNO; | |
118e2215 | 5524 | assert(0); /* NOTREACHED */ |
3c0563b9 | 5525 | |
5d458dd8 YO |
5526 | case BRANCH_next: |
5527 | sayYES; | |
118e2215 | 5528 | assert(0); /* NOTREACHED */ |
3c0563b9 | 5529 | |
40a82448 | 5530 | case BRANCH_next_fail: /* that branch failed; try the next, if any */ |
5d458dd8 YO |
5531 | if (do_cutgroup) { |
5532 | do_cutgroup = 0; | |
5533 | no_final = 0; | |
5534 | } | |
40a82448 | 5535 | REGCP_UNWIND(ST.cp); |
a8d1f4b4 | 5536 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); |
40a82448 DM |
5537 | scan = ST.next_branch; |
5538 | /* no more branches? */ | |
5d458dd8 YO |
5539 | if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) { |
5540 | DEBUG_EXECUTE_r({ | |
5541 | PerlIO_printf( Perl_debug_log, | |
5542 | "%*s %sBRANCH failed...%s\n", | |
5543 | REPORT_CODE_OFF+depth*2, "", | |
5544 | PL_colors[4], | |
5545 | PL_colors[5] ); | |
5546 | }); | |
5547 | sayNO_SILENT; | |
5548 | } | |
40a82448 | 5549 | continue; /* execute next BRANCH[J] op */ |
118e2215 | 5550 | assert(0); /* NOTREACHED */ |
40a82448 | 5551 | |
3c0563b9 | 5552 | case MINMOD: /* next op will be non-greedy, e.g. A*? */ |
24d3c4a9 | 5553 | minmod = 1; |
a0d0e21e | 5554 | break; |
40a82448 DM |
5555 | |
5556 | #undef ST | |
5557 | #define ST st->u.curlym | |
5558 | ||
5559 | case CURLYM: /* /A{m,n}B/ where A is fixed-length */ | |
5560 | ||
5561 | /* This is an optimisation of CURLYX that enables us to push | |
84d2fa14 | 5562 | * only a single backtracking state, no matter how many matches |
40a82448 DM |
5563 | * there are in {m,n}. It relies on the pattern being constant |
5564 | * length, with no parens to influence future backrefs | |
5565 | */ | |
5566 | ||
5567 | ST.me = scan; | |
dc45a647 | 5568 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; |
40a82448 | 5569 | |
f6033a9d DM |
5570 | ST.lastparen = rex->lastparen; |
5571 | ST.lastcloseparen = rex->lastcloseparen; | |
5572 | ||
40a82448 DM |
5573 | /* if paren positive, emulate an OPEN/CLOSE around A */ |
5574 | if (ST.me->flags) { | |
3b6647e0 | 5575 | U32 paren = ST.me->flags; |
40a82448 DM |
5576 | if (paren > PL_regsize) |
5577 | PL_regsize = paren; | |
c277df42 | 5578 | scan += NEXT_OFF(scan); /* Skip former OPEN. */ |
6407bf3b | 5579 | } |
40a82448 DM |
5580 | ST.A = scan; |
5581 | ST.B = next; | |
5582 | ST.alen = 0; | |
5583 | ST.count = 0; | |
24d3c4a9 DM |
5584 | ST.minmod = minmod; |
5585 | minmod = 0; | |
40a82448 DM |
5586 | ST.c1 = CHRTEST_UNINIT; |
5587 | REGCP_SET(ST.cp); | |
6407bf3b | 5588 | |
40a82448 DM |
5589 | if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */ |
5590 | goto curlym_do_B; | |
5591 | ||
5592 | curlym_do_A: /* execute the A in /A{m,n}B/ */ | |
4d5016e5 | 5593 | PUSH_YES_STATE_GOTO(CURLYM_A, ST.A, locinput); /* match A */ |
118e2215 | 5594 | assert(0); /* NOTREACHED */ |
5f80c4cf | 5595 | |
40a82448 | 5596 | case CURLYM_A: /* we've just matched an A */ |
40a82448 DM |
5597 | ST.count++; |
5598 | /* after first match, determine A's length: u.curlym.alen */ | |
5599 | if (ST.count == 1) { | |
5600 | if (PL_reg_match_utf8) { | |
c07e9d7b DM |
5601 | char *s = st->locinput; |
5602 | while (s < locinput) { | |
40a82448 DM |
5603 | ST.alen++; |
5604 | s += UTF8SKIP(s); | |
5605 | } | |
5606 | } | |
5607 | else { | |
c07e9d7b | 5608 | ST.alen = locinput - st->locinput; |
40a82448 DM |
5609 | } |
5610 | if (ST.alen == 0) | |
5611 | ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me); | |
5612 | } | |
0cadcf80 DM |
5613 | DEBUG_EXECUTE_r( |
5614 | PerlIO_printf(Perl_debug_log, | |
40a82448 | 5615 | "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n", |
5bc10b2c | 5616 | (int)(REPORT_CODE_OFF+(depth*2)), "", |
40a82448 | 5617 | (IV) ST.count, (IV)ST.alen) |
0cadcf80 DM |
5618 | ); |
5619 | ||
0a4db386 | 5620 | if (cur_eval && cur_eval->u.eval.close_paren && |
24b23f37 | 5621 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
5622 | goto fake_end; |
5623 | ||
c966426a DM |
5624 | { |
5625 | I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)); | |
5626 | if ( max == REG_INFTY || ST.count < max ) | |
5627 | goto curlym_do_A; /* try to match another A */ | |
5628 | } | |
40a82448 | 5629 | goto curlym_do_B; /* try to match B */ |
5f80c4cf | 5630 | |
40a82448 DM |
5631 | case CURLYM_A_fail: /* just failed to match an A */ |
5632 | REGCP_UNWIND(ST.cp); | |
0a4db386 YO |
5633 | |
5634 | if (ST.minmod || ST.count < ARG1(ST.me) /* min*/ | |
5635 | || (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 5636 | cur_eval->u.eval.close_paren == (U32)ST.me->flags)) |
40a82448 | 5637 | sayNO; |
0cadcf80 | 5638 | |
40a82448 | 5639 | curlym_do_B: /* execute the B in /A{m,n}B/ */ |
40a82448 DM |
5640 | if (ST.c1 == CHRTEST_UNINIT) { |
5641 | /* calculate c1 and c2 for possible match of 1st char | |
5642 | * following curly */ | |
5643 | ST.c1 = ST.c2 = CHRTEST_VOID; | |
5644 | if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) { | |
5645 | regnode *text_node = ST.B; | |
5646 | if (! HAS_TEXT(text_node)) | |
5647 | FIND_NEXT_IMPT(text_node); | |
ee9b8eae YO |
5648 | /* this used to be |
5649 | ||
5650 | (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT) | |
5651 | ||
5652 | But the former is redundant in light of the latter. | |
5653 | ||
5654 | if this changes back then the macro for | |
5655 | IS_TEXT and friends need to change. | |
5656 | */ | |
c74f6de9 | 5657 | if (PL_regkind[OP(text_node)] == EXACT) { |
79a2a0e8 KW |
5658 | if (! S_setup_EXACTISH_ST_c1_c2(aTHX_ |
5659 | text_node, &ST.c1, ST.c1_utf8, &ST.c2, ST.c2_utf8)) | |
c74f6de9 KW |
5660 | { |
5661 | sayNO; | |
5662 | } | |
c277df42 | 5663 | } |
c277df42 | 5664 | } |
40a82448 DM |
5665 | } |
5666 | ||
5667 | DEBUG_EXECUTE_r( | |
5668 | PerlIO_printf(Perl_debug_log, | |
5669 | "%*s CURLYM trying tail with matches=%"IVdf"...\n", | |
5bc10b2c | 5670 | (int)(REPORT_CODE_OFF+(depth*2)), |
40a82448 DM |
5671 | "", (IV)ST.count) |
5672 | ); | |
c74f6de9 | 5673 | if (! NEXTCHR_IS_EOS && ST.c1 != CHRTEST_VOID) { |
79a2a0e8 KW |
5674 | if (! UTF8_IS_INVARIANT(nextchr) && utf8_target) { |
5675 | if (memNE(locinput, ST.c1_utf8, UTF8SKIP(locinput)) | |
5676 | && memNE(locinput, ST.c2_utf8, UTF8SKIP(locinput))) | |
5677 | { | |
5678 | /* simulate B failing */ | |
5679 | DEBUG_OPTIMISE_r( | |
5680 | PerlIO_printf(Perl_debug_log, | |
5681 | "%*s CURLYM Fast bail next target=U+%"UVXf" c1=U+%"UVXf" c2=U+%"UVXf"\n", | |
5682 | (int)(REPORT_CODE_OFF+(depth*2)),"", | |
5683 | valid_utf8_to_uvchr((U8 *) locinput, NULL), | |
5684 | valid_utf8_to_uvchr(ST.c1_utf8, NULL), | |
5685 | valid_utf8_to_uvchr(ST.c2_utf8, NULL)) | |
5686 | ); | |
5687 | state_num = CURLYM_B_fail; | |
5688 | goto reenter_switch; | |
5689 | } | |
5690 | } | |
5691 | else if (nextchr != ST.c1 && nextchr != ST.c2) { | |
5400f398 KW |
5692 | /* simulate B failing */ |
5693 | DEBUG_OPTIMISE_r( | |
5694 | PerlIO_printf(Perl_debug_log, | |
79a2a0e8 | 5695 | "%*s CURLYM Fast bail next target=U+%X c1=U+%X c2=U+%X\n", |
5400f398 | 5696 | (int)(REPORT_CODE_OFF+(depth*2)),"", |
79a2a0e8 KW |
5697 | (int) nextchr, ST.c1, ST.c2) |
5698 | ); | |
5400f398 KW |
5699 | state_num = CURLYM_B_fail; |
5700 | goto reenter_switch; | |
5701 | } | |
c74f6de9 | 5702 | } |
40a82448 DM |
5703 | |
5704 | if (ST.me->flags) { | |
f6033a9d | 5705 | /* emulate CLOSE: mark current A as captured */ |
40a82448 DM |
5706 | I32 paren = ST.me->flags; |
5707 | if (ST.count) { | |
b93070ed | 5708 | rex->offs[paren].start |
c07e9d7b DM |
5709 | = HOPc(locinput, -ST.alen) - PL_bostr; |
5710 | rex->offs[paren].end = locinput - PL_bostr; | |
f6033a9d DM |
5711 | if ((U32)paren > rex->lastparen) |
5712 | rex->lastparen = paren; | |
5713 | rex->lastcloseparen = paren; | |
c277df42 | 5714 | } |
40a82448 | 5715 | else |
b93070ed | 5716 | rex->offs[paren].end = -1; |
0a4db386 | 5717 | if (cur_eval && cur_eval->u.eval.close_paren && |
24b23f37 | 5718 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
5719 | { |
5720 | if (ST.count) | |
5721 | goto fake_end; | |
5722 | else | |
5723 | sayNO; | |
5724 | } | |
c277df42 | 5725 | } |
0a4db386 | 5726 | |
4d5016e5 | 5727 | PUSH_STATE_GOTO(CURLYM_B, ST.B, locinput); /* match B */ |
118e2215 | 5728 | assert(0); /* NOTREACHED */ |
40a82448 DM |
5729 | |
5730 | case CURLYM_B_fail: /* just failed to match a B */ | |
5731 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 | 5732 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); |
40a82448 | 5733 | if (ST.minmod) { |
84d2fa14 HS |
5734 | I32 max = ARG2(ST.me); |
5735 | if (max != REG_INFTY && ST.count == max) | |
40a82448 DM |
5736 | sayNO; |
5737 | goto curlym_do_A; /* try to match a further A */ | |
5738 | } | |
5739 | /* backtrack one A */ | |
5740 | if (ST.count == ARG1(ST.me) /* min */) | |
5741 | sayNO; | |
5742 | ST.count--; | |
7016d6eb | 5743 | SET_locinput(HOPc(locinput, -ST.alen)); |
40a82448 DM |
5744 | goto curlym_do_B; /* try to match B */ |
5745 | ||
c255a977 DM |
5746 | #undef ST |
5747 | #define ST st->u.curly | |
40a82448 | 5748 | |
c255a977 DM |
5749 | #define CURLY_SETPAREN(paren, success) \ |
5750 | if (paren) { \ | |
5751 | if (success) { \ | |
b93070ed DM |
5752 | rex->offs[paren].start = HOPc(locinput, -1) - PL_bostr; \ |
5753 | rex->offs[paren].end = locinput - PL_bostr; \ | |
f6033a9d DM |
5754 | if (paren > rex->lastparen) \ |
5755 | rex->lastparen = paren; \ | |
b93070ed | 5756 | rex->lastcloseparen = paren; \ |
c255a977 | 5757 | } \ |
f6033a9d | 5758 | else { \ |
b93070ed | 5759 | rex->offs[paren].end = -1; \ |
f6033a9d DM |
5760 | rex->lastparen = ST.lastparen; \ |
5761 | rex->lastcloseparen = ST.lastcloseparen; \ | |
5762 | } \ | |
c255a977 DM |
5763 | } |
5764 | ||
b40a2c17 | 5765 | case STAR: /* /A*B/ where A is width 1 char */ |
c255a977 DM |
5766 | ST.paren = 0; |
5767 | ST.min = 0; | |
5768 | ST.max = REG_INFTY; | |
a0d0e21e LW |
5769 | scan = NEXTOPER(scan); |
5770 | goto repeat; | |
3c0563b9 | 5771 | |
b40a2c17 | 5772 | case PLUS: /* /A+B/ where A is width 1 char */ |
c255a977 DM |
5773 | ST.paren = 0; |
5774 | ST.min = 1; | |
5775 | ST.max = REG_INFTY; | |
c277df42 | 5776 | scan = NEXTOPER(scan); |
c255a977 | 5777 | goto repeat; |
3c0563b9 | 5778 | |
b40a2c17 | 5779 | case CURLYN: /* /(A){m,n}B/ where A is width 1 char */ |
5400f398 KW |
5780 | ST.paren = scan->flags; /* Which paren to set */ |
5781 | ST.lastparen = rex->lastparen; | |
f6033a9d | 5782 | ST.lastcloseparen = rex->lastcloseparen; |
c255a977 DM |
5783 | if (ST.paren > PL_regsize) |
5784 | PL_regsize = ST.paren; | |
c255a977 DM |
5785 | ST.min = ARG1(scan); /* min to match */ |
5786 | ST.max = ARG2(scan); /* max to match */ | |
0a4db386 | 5787 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5788 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5789 | ST.min=1; |
5790 | ST.max=1; | |
5791 | } | |
c255a977 DM |
5792 | scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE); |
5793 | goto repeat; | |
3c0563b9 | 5794 | |
b40a2c17 | 5795 | case CURLY: /* /A{m,n}B/ where A is width 1 char */ |
c255a977 DM |
5796 | ST.paren = 0; |
5797 | ST.min = ARG1(scan); /* min to match */ | |
5798 | ST.max = ARG2(scan); /* max to match */ | |
5799 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; | |
c277df42 | 5800 | repeat: |
a0d0e21e LW |
5801 | /* |
5802 | * Lookahead to avoid useless match attempts | |
5803 | * when we know what character comes next. | |
c255a977 | 5804 | * |
5f80c4cf JP |
5805 | * Used to only do .*x and .*?x, but now it allows |
5806 | * for )'s, ('s and (?{ ... })'s to be in the way | |
5807 | * of the quantifier and the EXACT-like node. -- japhy | |
5808 | */ | |
5809 | ||
eb5c1be8 | 5810 | assert(ST.min <= ST.max); |
3337dfe3 KW |
5811 | if (! HAS_TEXT(next) && ! JUMPABLE(next)) { |
5812 | ST.c1 = ST.c2 = CHRTEST_VOID; | |
5813 | } | |
5814 | else { | |
5f80c4cf JP |
5815 | regnode *text_node = next; |
5816 | ||
3dab1dad YO |
5817 | if (! HAS_TEXT(text_node)) |
5818 | FIND_NEXT_IMPT(text_node); | |
5f80c4cf | 5819 | |
9e137952 | 5820 | if (! HAS_TEXT(text_node)) |
c255a977 | 5821 | ST.c1 = ST.c2 = CHRTEST_VOID; |
5f80c4cf | 5822 | else { |
ee9b8eae | 5823 | if ( PL_regkind[OP(text_node)] != EXACT ) { |
c255a977 | 5824 | ST.c1 = ST.c2 = CHRTEST_VOID; |
cca55fe3 | 5825 | } |
c74f6de9 | 5826 | else { |
ee9b8eae YO |
5827 | |
5828 | /* Currently we only get here when | |
5829 | ||
5830 | PL_rekind[OP(text_node)] == EXACT | |
5831 | ||
5832 | if this changes back then the macro for IS_TEXT and | |
5833 | friends need to change. */ | |
79a2a0e8 KW |
5834 | if (! S_setup_EXACTISH_ST_c1_c2(aTHX_ |
5835 | text_node, &ST.c1, ST.c1_utf8, &ST.c2, ST.c2_utf8)) | |
c74f6de9 KW |
5836 | { |
5837 | sayNO; | |
5838 | } | |
5839 | } | |
1aa99e6b | 5840 | } |
bbce6d69 | 5841 | } |
c255a977 DM |
5842 | |
5843 | ST.A = scan; | |
5844 | ST.B = next; | |
24d3c4a9 | 5845 | if (minmod) { |
eb72505d | 5846 | char *li = locinput; |
24d3c4a9 | 5847 | minmod = 0; |
eb72505d | 5848 | if (ST.min && regrepeat(rex, &li, ST.A, ST.min, depth) < ST.min) |
4633a7c4 | 5849 | sayNO; |
7016d6eb | 5850 | SET_locinput(li); |
c255a977 | 5851 | ST.count = ST.min; |
c255a977 DM |
5852 | REGCP_SET(ST.cp); |
5853 | if (ST.c1 == CHRTEST_VOID) | |
5854 | goto curly_try_B_min; | |
5855 | ||
5856 | ST.oldloc = locinput; | |
5857 | ||
5858 | /* set ST.maxpos to the furthest point along the | |
5859 | * string that could possibly match */ | |
5860 | if (ST.max == REG_INFTY) { | |
5861 | ST.maxpos = PL_regeol - 1; | |
f2ed9b32 | 5862 | if (utf8_target) |
c255a977 DM |
5863 | while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos)) |
5864 | ST.maxpos--; | |
5865 | } | |
f2ed9b32 | 5866 | else if (utf8_target) { |
c255a977 DM |
5867 | int m = ST.max - ST.min; |
5868 | for (ST.maxpos = locinput; | |
5869 | m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--) | |
5870 | ST.maxpos += UTF8SKIP(ST.maxpos); | |
5871 | } | |
5872 | else { | |
5873 | ST.maxpos = locinput + ST.max - ST.min; | |
5874 | if (ST.maxpos >= PL_regeol) | |
5875 | ST.maxpos = PL_regeol - 1; | |
5876 | } | |
5877 | goto curly_try_B_min_known; | |
5878 | ||
5879 | } | |
5880 | else { | |
eb72505d DM |
5881 | /* avoid taking address of locinput, so it can remain |
5882 | * a register var */ | |
5883 | char *li = locinput; | |
5884 | ST.count = regrepeat(rex, &li, ST.A, ST.max, depth); | |
c255a977 DM |
5885 | if (ST.count < ST.min) |
5886 | sayNO; | |
7016d6eb | 5887 | SET_locinput(li); |
c255a977 DM |
5888 | if ((ST.count > ST.min) |
5889 | && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL)) | |
5890 | { | |
5891 | /* A{m,n} must come at the end of the string, there's | |
5892 | * no point in backing off ... */ | |
5893 | ST.min = ST.count; | |
5894 | /* ...except that $ and \Z can match before *and* after | |
5895 | newline at the end. Consider "\n\n" =~ /\n+\Z\n/. | |
5896 | We may back off by one in this case. */ | |
eb72505d | 5897 | if (UCHARAT(locinput - 1) == '\n' && OP(ST.B) != EOS) |
c255a977 DM |
5898 | ST.min--; |
5899 | } | |
5900 | REGCP_SET(ST.cp); | |
5901 | goto curly_try_B_max; | |
5902 | } | |
118e2215 | 5903 | assert(0); /* NOTREACHED */ |
c255a977 DM |
5904 | |
5905 | ||
5906 | case CURLY_B_min_known_fail: | |
5907 | /* failed to find B in a non-greedy match where c1,c2 valid */ | |
c255a977 | 5908 | |
c255a977 | 5909 | REGCP_UNWIND(ST.cp); |
a8d1f4b4 DM |
5910 | if (ST.paren) { |
5911 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); | |
5912 | } | |
c255a977 DM |
5913 | /* Couldn't or didn't -- move forward. */ |
5914 | ST.oldloc = locinput; | |
f2ed9b32 | 5915 | if (utf8_target) |
c255a977 DM |
5916 | locinput += UTF8SKIP(locinput); |
5917 | else | |
5918 | locinput++; | |
5919 | ST.count++; | |
5920 | curly_try_B_min_known: | |
5921 | /* find the next place where 'B' could work, then call B */ | |
5922 | { | |
5923 | int n; | |
f2ed9b32 | 5924 | if (utf8_target) { |
c255a977 DM |
5925 | n = (ST.oldloc == locinput) ? 0 : 1; |
5926 | if (ST.c1 == ST.c2) { | |
c255a977 | 5927 | /* set n to utf8_distance(oldloc, locinput) */ |
79a2a0e8 KW |
5928 | while (locinput <= ST.maxpos |
5929 | && memNE(locinput, ST.c1_utf8, UTF8SKIP(locinput))) | |
5930 | { | |
5931 | locinput += UTF8SKIP(locinput); | |
c255a977 DM |
5932 | n++; |
5933 | } | |
1aa99e6b IH |
5934 | } |
5935 | else { | |
c255a977 | 5936 | /* set n to utf8_distance(oldloc, locinput) */ |
79a2a0e8 KW |
5937 | while (locinput <= ST.maxpos |
5938 | && memNE(locinput, ST.c1_utf8, UTF8SKIP(locinput)) | |
5939 | && memNE(locinput, ST.c2_utf8, UTF8SKIP(locinput))) | |
5940 | { | |
5941 | locinput += UTF8SKIP(locinput); | |
c255a977 | 5942 | n++; |
1aa99e6b | 5943 | } |
0fe9bf95 IZ |
5944 | } |
5945 | } | |
5400f398 | 5946 | else { /* Not utf8_target */ |
c255a977 DM |
5947 | if (ST.c1 == ST.c2) { |
5948 | while (locinput <= ST.maxpos && | |
5949 | UCHARAT(locinput) != ST.c1) | |
5950 | locinput++; | |
bbce6d69 | 5951 | } |
c255a977 DM |
5952 | else { |
5953 | while (locinput <= ST.maxpos | |
5954 | && UCHARAT(locinput) != ST.c1 | |
5955 | && UCHARAT(locinput) != ST.c2) | |
5956 | locinput++; | |
a0ed51b3 | 5957 | } |
c255a977 DM |
5958 | n = locinput - ST.oldloc; |
5959 | } | |
5960 | if (locinput > ST.maxpos) | |
5961 | sayNO; | |
c255a977 | 5962 | if (n) { |
eb72505d DM |
5963 | /* In /a{m,n}b/, ST.oldloc is at "a" x m, locinput is |
5964 | * at b; check that everything between oldloc and | |
5965 | * locinput matches */ | |
5966 | char *li = ST.oldloc; | |
c255a977 | 5967 | ST.count += n; |
eb72505d | 5968 | if (regrepeat(rex, &li, ST.A, n, depth) < n) |
4633a7c4 | 5969 | sayNO; |
eb72505d | 5970 | assert(n == REG_INFTY || locinput == li); |
a0d0e21e | 5971 | } |
c255a977 | 5972 | CURLY_SETPAREN(ST.paren, ST.count); |
0a4db386 | 5973 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5974 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5975 | goto fake_end; |
5976 | } | |
4d5016e5 | 5977 | PUSH_STATE_GOTO(CURLY_B_min_known, ST.B, locinput); |
a0d0e21e | 5978 | } |
118e2215 | 5979 | assert(0); /* NOTREACHED */ |
c255a977 DM |
5980 | |
5981 | ||
5982 | case CURLY_B_min_fail: | |
5983 | /* failed to find B in a non-greedy match where c1,c2 invalid */ | |
c255a977 DM |
5984 | |
5985 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 DM |
5986 | if (ST.paren) { |
5987 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); | |
5988 | } | |
c255a977 | 5989 | /* failed -- move forward one */ |
f73aaa43 | 5990 | { |
eb72505d DM |
5991 | char *li = locinput; |
5992 | if (!regrepeat(rex, &li, ST.A, 1, depth)) { | |
f73aaa43 DM |
5993 | sayNO; |
5994 | } | |
eb72505d | 5995 | locinput = li; |
f73aaa43 DM |
5996 | } |
5997 | { | |
c255a977 | 5998 | ST.count++; |
c255a977 DM |
5999 | if (ST.count <= ST.max || (ST.max == REG_INFTY && |
6000 | ST.count > 0)) /* count overflow ? */ | |
15272685 | 6001 | { |
c255a977 DM |
6002 | curly_try_B_min: |
6003 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 6004 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 6005 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
6006 | goto fake_end; |
6007 | } | |
4d5016e5 | 6008 | PUSH_STATE_GOTO(CURLY_B_min, ST.B, locinput); |
a0d0e21e LW |
6009 | } |
6010 | } | |
c74f6de9 | 6011 | sayNO; |
118e2215 | 6012 | assert(0); /* NOTREACHED */ |
c255a977 DM |
6013 | |
6014 | ||
6015 | curly_try_B_max: | |
6016 | /* a successful greedy match: now try to match B */ | |
40d049e4 | 6017 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 6018 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
40d049e4 YO |
6019 | goto fake_end; |
6020 | } | |
c255a977 | 6021 | { |
79a2a0e8 KW |
6022 | bool could_match = locinput < PL_regeol; |
6023 | ||
c255a977 | 6024 | /* If it could work, try it. */ |
79a2a0e8 KW |
6025 | if (ST.c1 != CHRTEST_VOID && could_match) { |
6026 | if (! UTF8_IS_INVARIANT(UCHARAT(locinput)) && utf8_target) | |
6027 | { | |
6028 | could_match = memEQ(locinput, | |
6029 | ST.c1_utf8, | |
6030 | UTF8SKIP(locinput)) | |
6031 | || memEQ(locinput, | |
6032 | ST.c2_utf8, | |
6033 | UTF8SKIP(locinput)); | |
6034 | } | |
6035 | else { | |
6036 | could_match = UCHARAT(locinput) == ST.c1 | |
6037 | || UCHARAT(locinput) == ST.c2; | |
6038 | } | |
6039 | } | |
6040 | if (ST.c1 == CHRTEST_VOID || could_match) { | |
c255a977 | 6041 | CURLY_SETPAREN(ST.paren, ST.count); |
4d5016e5 | 6042 | PUSH_STATE_GOTO(CURLY_B_max, ST.B, locinput); |
118e2215 | 6043 | assert(0); /* NOTREACHED */ |
c255a977 DM |
6044 | } |
6045 | } | |
6046 | /* FALL THROUGH */ | |
3c0563b9 | 6047 | |
c255a977 DM |
6048 | case CURLY_B_max_fail: |
6049 | /* failed to find B in a greedy match */ | |
c255a977 DM |
6050 | |
6051 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 DM |
6052 | if (ST.paren) { |
6053 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); | |
6054 | } | |
c255a977 DM |
6055 | /* back up. */ |
6056 | if (--ST.count < ST.min) | |
6057 | sayNO; | |
eb72505d | 6058 | locinput = HOPc(locinput, -1); |
c255a977 DM |
6059 | goto curly_try_B_max; |
6060 | ||
6061 | #undef ST | |
6062 | ||
3c0563b9 | 6063 | case END: /* last op of main pattern */ |
6bda09f9 | 6064 | fake_end: |
faec1544 DM |
6065 | if (cur_eval) { |
6066 | /* we've just finished A in /(??{A})B/; now continue with B */ | |
faec1544 DM |
6067 | st->u.eval.toggle_reg_flags |
6068 | = cur_eval->u.eval.toggle_reg_flags; | |
6069 | PL_reg_flags ^= st->u.eval.toggle_reg_flags; | |
6070 | ||
288b8c02 | 6071 | st->u.eval.prev_rex = rex_sv; /* inner */ |
b93070ed | 6072 | st->u.eval.cp = regcppush(rex, 0); /* Save *all* the positions. */ |
ec43f78b DM |
6073 | rex_sv = cur_eval->u.eval.prev_rex; |
6074 | SET_reg_curpm(rex_sv); | |
288b8c02 | 6075 | rex = (struct regexp *)SvANY(rex_sv); |
f8fc2ecf | 6076 | rexi = RXi_GET(rex); |
faec1544 | 6077 | cur_curlyx = cur_eval->u.eval.prev_curlyx; |
34a81e2b | 6078 | |
faec1544 | 6079 | REGCP_SET(st->u.eval.lastcp); |
faec1544 DM |
6080 | |
6081 | /* Restore parens of the outer rex without popping the | |
6082 | * savestack */ | |
74088413 | 6083 | S_regcp_restore(aTHX_ rex, cur_eval->u.eval.lastcp); |
faec1544 DM |
6084 | |
6085 | st->u.eval.prev_eval = cur_eval; | |
6086 | cur_eval = cur_eval->u.eval.prev_eval; | |
6087 | DEBUG_EXECUTE_r( | |
2a49f0f5 JH |
6088 | PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n", |
6089 | REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval));); | |
e7707071 YO |
6090 | if ( nochange_depth ) |
6091 | nochange_depth--; | |
6092 | ||
4d5016e5 DM |
6093 | PUSH_YES_STATE_GOTO(EVAL_AB, st->u.eval.prev_eval->u.eval.B, |
6094 | locinput); /* match B */ | |
faec1544 DM |
6095 | } |
6096 | ||
3b0527fe | 6097 | if (locinput < reginfo->till) { |
a3621e74 | 6098 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
7821416a IZ |
6099 | "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n", |
6100 | PL_colors[4], | |
6101 | (long)(locinput - PL_reg_starttry), | |
3b0527fe | 6102 | (long)(reginfo->till - PL_reg_starttry), |
7821416a | 6103 | PL_colors[5])); |
58e23c8d | 6104 | |
262b90c4 | 6105 | sayNO_SILENT; /* Cannot match: too short. */ |
7821416a | 6106 | } |
262b90c4 | 6107 | sayYES; /* Success! */ |
dad79028 DM |
6108 | |
6109 | case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */ | |
6110 | DEBUG_EXECUTE_r( | |
6111 | PerlIO_printf(Perl_debug_log, | |
6112 | "%*s %ssubpattern success...%s\n", | |
5bc10b2c | 6113 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])); |
262b90c4 | 6114 | sayYES; /* Success! */ |
dad79028 | 6115 | |
40a82448 DM |
6116 | #undef ST |
6117 | #define ST st->u.ifmatch | |
6118 | ||
37f53970 DM |
6119 | { |
6120 | char *newstart; | |
6121 | ||
40a82448 DM |
6122 | case SUSPEND: /* (?>A) */ |
6123 | ST.wanted = 1; | |
37f53970 | 6124 | newstart = locinput; |
9041c2e3 | 6125 | goto do_ifmatch; |
dad79028 | 6126 | |
40a82448 DM |
6127 | case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */ |
6128 | ST.wanted = 0; | |
dad79028 DM |
6129 | goto ifmatch_trivial_fail_test; |
6130 | ||
40a82448 DM |
6131 | case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */ |
6132 | ST.wanted = 1; | |
dad79028 | 6133 | ifmatch_trivial_fail_test: |
a0ed51b3 | 6134 | if (scan->flags) { |
52657f30 | 6135 | char * const s = HOPBACKc(locinput, scan->flags); |
dad79028 DM |
6136 | if (!s) { |
6137 | /* trivial fail */ | |
24d3c4a9 DM |
6138 | if (logical) { |
6139 | logical = 0; | |
f2338a2e | 6140 | sw = 1 - cBOOL(ST.wanted); |
dad79028 | 6141 | } |
40a82448 | 6142 | else if (ST.wanted) |
dad79028 DM |
6143 | sayNO; |
6144 | next = scan + ARG(scan); | |
6145 | if (next == scan) | |
6146 | next = NULL; | |
6147 | break; | |
6148 | } | |
37f53970 | 6149 | newstart = s; |
a0ed51b3 LW |
6150 | } |
6151 | else | |
37f53970 | 6152 | newstart = locinput; |
a0ed51b3 | 6153 | |
c277df42 | 6154 | do_ifmatch: |
40a82448 | 6155 | ST.me = scan; |
24d3c4a9 | 6156 | ST.logical = logical; |
24d786f4 YO |
6157 | logical = 0; /* XXX: reset state of logical once it has been saved into ST */ |
6158 | ||
40a82448 | 6159 | /* execute body of (?...A) */ |
37f53970 | 6160 | PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan)), newstart); |
118e2215 | 6161 | assert(0); /* NOTREACHED */ |
37f53970 | 6162 | } |
40a82448 DM |
6163 | |
6164 | case IFMATCH_A_fail: /* body of (?...A) failed */ | |
6165 | ST.wanted = !ST.wanted; | |
6166 | /* FALL THROUGH */ | |
6167 | ||
6168 | case IFMATCH_A: /* body of (?...A) succeeded */ | |
24d3c4a9 | 6169 | if (ST.logical) { |
f2338a2e | 6170 | sw = cBOOL(ST.wanted); |
40a82448 DM |
6171 | } |
6172 | else if (!ST.wanted) | |
6173 | sayNO; | |
6174 | ||
37f53970 DM |
6175 | if (OP(ST.me) != SUSPEND) { |
6176 | /* restore old position except for (?>...) */ | |
6177 | locinput = st->locinput; | |
40a82448 DM |
6178 | } |
6179 | scan = ST.me + ARG(ST.me); | |
6180 | if (scan == ST.me) | |
6181 | scan = NULL; | |
6182 | continue; /* execute B */ | |
6183 | ||
6184 | #undef ST | |
dad79028 | 6185 | |
3c0563b9 DM |
6186 | case LONGJMP: /* alternative with many branches compiles to |
6187 | * (BRANCHJ; EXACT ...; LONGJMP ) x N */ | |
c277df42 IZ |
6188 | next = scan + ARG(scan); |
6189 | if (next == scan) | |
6190 | next = NULL; | |
a0d0e21e | 6191 | break; |
3c0563b9 DM |
6192 | |
6193 | case COMMIT: /* (*COMMIT) */ | |
e2e6a0f1 YO |
6194 | reginfo->cutpoint = PL_regeol; |
6195 | /* FALLTHROUGH */ | |
3c0563b9 DM |
6196 | |
6197 | case PRUNE: /* (*PRUNE) */ | |
e2e6a0f1 | 6198 | if (!scan->flags) |
ad64d0ec | 6199 | sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
4d5016e5 | 6200 | PUSH_STATE_GOTO(COMMIT_next, next, locinput); |
118e2215 | 6201 | assert(0); /* NOTREACHED */ |
3c0563b9 | 6202 | |
54612592 YO |
6203 | case COMMIT_next_fail: |
6204 | no_final = 1; | |
6205 | /* FALLTHROUGH */ | |
3c0563b9 DM |
6206 | |
6207 | case OPFAIL: /* (*FAIL) */ | |
7f69552c | 6208 | sayNO; |
118e2215 | 6209 | assert(0); /* NOTREACHED */ |
e2e6a0f1 YO |
6210 | |
6211 | #define ST st->u.mark | |
3c0563b9 | 6212 | case MARKPOINT: /* (*MARK:foo) */ |
e2e6a0f1 | 6213 | ST.prev_mark = mark_state; |
5d458dd8 | 6214 | ST.mark_name = sv_commit = sv_yes_mark |
ad64d0ec | 6215 | = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
e2e6a0f1 | 6216 | mark_state = st; |
4d5016e5 DM |
6217 | ST.mark_loc = locinput; |
6218 | PUSH_YES_STATE_GOTO(MARKPOINT_next, next, locinput); | |
118e2215 | 6219 | assert(0); /* NOTREACHED */ |
3c0563b9 | 6220 | |
e2e6a0f1 YO |
6221 | case MARKPOINT_next: |
6222 | mark_state = ST.prev_mark; | |
6223 | sayYES; | |
118e2215 | 6224 | assert(0); /* NOTREACHED */ |
3c0563b9 | 6225 | |
e2e6a0f1 | 6226 | case MARKPOINT_next_fail: |
5d458dd8 | 6227 | if (popmark && sv_eq(ST.mark_name,popmark)) |
e2e6a0f1 YO |
6228 | { |
6229 | if (ST.mark_loc > startpoint) | |
6230 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
6231 | popmark = NULL; /* we found our mark */ | |
6232 | sv_commit = ST.mark_name; | |
6233 | ||
6234 | DEBUG_EXECUTE_r({ | |
5d458dd8 | 6235 | PerlIO_printf(Perl_debug_log, |
e2e6a0f1 YO |
6236 | "%*s %ssetting cutpoint to mark:%"SVf"...%s\n", |
6237 | REPORT_CODE_OFF+depth*2, "", | |
be2597df | 6238 | PL_colors[4], SVfARG(sv_commit), PL_colors[5]); |
e2e6a0f1 YO |
6239 | }); |
6240 | } | |
6241 | mark_state = ST.prev_mark; | |
5d458dd8 YO |
6242 | sv_yes_mark = mark_state ? |
6243 | mark_state->u.mark.mark_name : NULL; | |
e2e6a0f1 | 6244 | sayNO; |
118e2215 | 6245 | assert(0); /* NOTREACHED */ |
3c0563b9 DM |
6246 | |
6247 | case SKIP: /* (*SKIP) */ | |
5d458dd8 | 6248 | if (scan->flags) { |
2bf803e2 | 6249 | /* (*SKIP) : if we fail we cut here*/ |
5d458dd8 | 6250 | ST.mark_name = NULL; |
e2e6a0f1 | 6251 | ST.mark_loc = locinput; |
4d5016e5 | 6252 | PUSH_STATE_GOTO(SKIP_next,next, locinput); |
5d458dd8 | 6253 | } else { |
2bf803e2 | 6254 | /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was, |
5d458dd8 YO |
6255 | otherwise do nothing. Meaning we need to scan |
6256 | */ | |
6257 | regmatch_state *cur = mark_state; | |
ad64d0ec | 6258 | SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
6259 | |
6260 | while (cur) { | |
6261 | if ( sv_eq( cur->u.mark.mark_name, | |
6262 | find ) ) | |
6263 | { | |
6264 | ST.mark_name = find; | |
4d5016e5 | 6265 | PUSH_STATE_GOTO( SKIP_next, next, locinput); |
5d458dd8 YO |
6266 | } |
6267 | cur = cur->u.mark.prev_mark; | |
6268 | } | |
e2e6a0f1 | 6269 | } |
2bf803e2 | 6270 | /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */ |
5d458dd8 | 6271 | break; |
3c0563b9 | 6272 | |
5d458dd8 YO |
6273 | case SKIP_next_fail: |
6274 | if (ST.mark_name) { | |
6275 | /* (*CUT:NAME) - Set up to search for the name as we | |
6276 | collapse the stack*/ | |
6277 | popmark = ST.mark_name; | |
6278 | } else { | |
6279 | /* (*CUT) - No name, we cut here.*/ | |
e2e6a0f1 YO |
6280 | if (ST.mark_loc > startpoint) |
6281 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5d458dd8 YO |
6282 | /* but we set sv_commit to latest mark_name if there |
6283 | is one so they can test to see how things lead to this | |
6284 | cut */ | |
6285 | if (mark_state) | |
6286 | sv_commit=mark_state->u.mark.mark_name; | |
6287 | } | |
e2e6a0f1 YO |
6288 | no_final = 1; |
6289 | sayNO; | |
118e2215 | 6290 | assert(0); /* NOTREACHED */ |
e2e6a0f1 | 6291 | #undef ST |
3c0563b9 DM |
6292 | |
6293 | case LNBREAK: /* \R */ | |
7016d6eb | 6294 | if ((n=is_LNBREAK_safe(locinput, PL_regeol, utf8_target))) { |
e1d1eefb | 6295 | locinput += n; |
e1d1eefb YO |
6296 | } else |
6297 | sayNO; | |
6298 | break; | |
6299 | ||
6300 | #define CASE_CLASS(nAmE) \ | |
6301 | case nAmE: \ | |
7016d6eb | 6302 | if (NEXTCHR_IS_EOS) \ |
1380b51d | 6303 | sayNO; \ |
f2ed9b32 | 6304 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb | 6305 | locinput += n; \ |
e1d1eefb YO |
6306 | } else \ |
6307 | sayNO; \ | |
6308 | break; \ | |
6309 | case N##nAmE: \ | |
7016d6eb | 6310 | if (NEXTCHR_IS_EOS) \ |
1380b51d | 6311 | sayNO; \ |
f2ed9b32 | 6312 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
6313 | sayNO; \ |
6314 | } else { \ | |
6315 | locinput += UTF8SKIP(locinput); \ | |
e1d1eefb YO |
6316 | } \ |
6317 | break | |
6318 | ||
3c0563b9 DM |
6319 | CASE_CLASS(VERTWS); /* \v \V */ |
6320 | CASE_CLASS(HORIZWS); /* \h \H */ | |
e1d1eefb YO |
6321 | #undef CASE_CLASS |
6322 | ||
a0d0e21e | 6323 | default: |
b900a521 | 6324 | PerlIO_printf(Perl_error_log, "%"UVxf" %d\n", |
d7d93a81 | 6325 | PTR2UV(scan), OP(scan)); |
cea2e8a9 | 6326 | Perl_croak(aTHX_ "regexp memory corruption"); |
28b98f76 DM |
6327 | |
6328 | /* this is a point to jump to in order to increment | |
6329 | * locinput by one character */ | |
6330 | increment_locinput: | |
e6ca698c | 6331 | assert(!NEXTCHR_IS_EOS); |
28b98f76 DM |
6332 | if (utf8_target) { |
6333 | locinput += PL_utf8skip[nextchr]; | |
7016d6eb | 6334 | /* locinput is allowed to go 1 char off the end, but not 2+ */ |
28b98f76 DM |
6335 | if (locinput > PL_regeol) |
6336 | sayNO; | |
28b98f76 DM |
6337 | } |
6338 | else | |
3640db6b | 6339 | locinput++; |
28b98f76 | 6340 | break; |
5d458dd8 YO |
6341 | |
6342 | } /* end switch */ | |
95b24440 | 6343 | |
5d458dd8 YO |
6344 | /* switch break jumps here */ |
6345 | scan = next; /* prepare to execute the next op and ... */ | |
6346 | continue; /* ... jump back to the top, reusing st */ | |
118e2215 | 6347 | assert(0); /* NOTREACHED */ |
95b24440 | 6348 | |
40a82448 DM |
6349 | push_yes_state: |
6350 | /* push a state that backtracks on success */ | |
6351 | st->u.yes.prev_yes_state = yes_state; | |
6352 | yes_state = st; | |
6353 | /* FALL THROUGH */ | |
6354 | push_state: | |
6355 | /* push a new regex state, then continue at scan */ | |
6356 | { | |
6357 | regmatch_state *newst; | |
6358 | ||
24b23f37 YO |
6359 | DEBUG_STACK_r({ |
6360 | regmatch_state *cur = st; | |
6361 | regmatch_state *curyes = yes_state; | |
6362 | int curd = depth; | |
6363 | regmatch_slab *slab = PL_regmatch_slab; | |
6364 | for (;curd > -1;cur--,curd--) { | |
6365 | if (cur < SLAB_FIRST(slab)) { | |
6366 | slab = slab->prev; | |
6367 | cur = SLAB_LAST(slab); | |
6368 | } | |
6369 | PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n", | |
6370 | REPORT_CODE_OFF + 2 + depth * 2,"", | |
13d6edb4 | 6371 | curd, PL_reg_name[cur->resume_state], |
24b23f37 YO |
6372 | (curyes == cur) ? "yes" : "" |
6373 | ); | |
6374 | if (curyes == cur) | |
6375 | curyes = cur->u.yes.prev_yes_state; | |
6376 | } | |
6377 | } else | |
6378 | DEBUG_STATE_pp("push") | |
6379 | ); | |
40a82448 | 6380 | depth++; |
40a82448 DM |
6381 | st->locinput = locinput; |
6382 | newst = st+1; | |
6383 | if (newst > SLAB_LAST(PL_regmatch_slab)) | |
6384 | newst = S_push_slab(aTHX); | |
6385 | PL_regmatch_state = newst; | |
786e8c11 | 6386 | |
4d5016e5 | 6387 | locinput = pushinput; |
40a82448 DM |
6388 | st = newst; |
6389 | continue; | |
118e2215 | 6390 | assert(0); /* NOTREACHED */ |
40a82448 | 6391 | } |
a0d0e21e | 6392 | } |
a687059c | 6393 | |
a0d0e21e LW |
6394 | /* |
6395 | * We get here only if there's trouble -- normally "case END" is | |
6396 | * the terminating point. | |
6397 | */ | |
cea2e8a9 | 6398 | Perl_croak(aTHX_ "corrupted regexp pointers"); |
a0d0e21e | 6399 | /*NOTREACHED*/ |
4633a7c4 LW |
6400 | sayNO; |
6401 | ||
262b90c4 | 6402 | yes: |
77cb431f DM |
6403 | if (yes_state) { |
6404 | /* we have successfully completed a subexpression, but we must now | |
6405 | * pop to the state marked by yes_state and continue from there */ | |
77cb431f | 6406 | assert(st != yes_state); |
5bc10b2c DM |
6407 | #ifdef DEBUGGING |
6408 | while (st != yes_state) { | |
6409 | st--; | |
6410 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
6411 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
6412 | st = SLAB_LAST(PL_regmatch_slab); | |
6413 | } | |
e2e6a0f1 | 6414 | DEBUG_STATE_r({ |
54612592 YO |
6415 | if (no_final) { |
6416 | DEBUG_STATE_pp("pop (no final)"); | |
6417 | } else { | |
6418 | DEBUG_STATE_pp("pop (yes)"); | |
6419 | } | |
e2e6a0f1 | 6420 | }); |
5bc10b2c DM |
6421 | depth--; |
6422 | } | |
6423 | #else | |
77cb431f DM |
6424 | while (yes_state < SLAB_FIRST(PL_regmatch_slab) |
6425 | || yes_state > SLAB_LAST(PL_regmatch_slab)) | |
6426 | { | |
6427 | /* not in this slab, pop slab */ | |
6428 | depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1); | |
6429 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
6430 | st = SLAB_LAST(PL_regmatch_slab); | |
6431 | } | |
6432 | depth -= (st - yes_state); | |
5bc10b2c | 6433 | #endif |
77cb431f DM |
6434 | st = yes_state; |
6435 | yes_state = st->u.yes.prev_yes_state; | |
6436 | PL_regmatch_state = st; | |
24b23f37 | 6437 | |
3640db6b | 6438 | if (no_final) |
5d458dd8 | 6439 | locinput= st->locinput; |
54612592 | 6440 | state_num = st->resume_state + no_final; |
24d3c4a9 | 6441 | goto reenter_switch; |
77cb431f DM |
6442 | } |
6443 | ||
a3621e74 | 6444 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n", |
e4584336 | 6445 | PL_colors[4], PL_colors[5])); |
02db2b7b | 6446 | |
ed301438 | 6447 | if (PL_reg_state.re_state_eval_setup_done) { |
19b95bf0 DM |
6448 | /* each successfully executed (?{...}) block does the equivalent of |
6449 | * local $^R = do {...} | |
6450 | * When popping the save stack, all these locals would be undone; | |
6451 | * bypass this by setting the outermost saved $^R to the latest | |
6452 | * value */ | |
6453 | if (oreplsv != GvSV(PL_replgv)) | |
6454 | sv_setsv(oreplsv, GvSV(PL_replgv)); | |
6455 | } | |
95b24440 | 6456 | result = 1; |
aa283a38 | 6457 | goto final_exit; |
4633a7c4 LW |
6458 | |
6459 | no: | |
a3621e74 | 6460 | DEBUG_EXECUTE_r( |
7821416a | 6461 | PerlIO_printf(Perl_debug_log, |
786e8c11 | 6462 | "%*s %sfailed...%s\n", |
5bc10b2c | 6463 | REPORT_CODE_OFF+depth*2, "", |
786e8c11 | 6464 | PL_colors[4], PL_colors[5]) |
7821416a | 6465 | ); |
aa283a38 | 6466 | |
262b90c4 | 6467 | no_silent: |
54612592 YO |
6468 | if (no_final) { |
6469 | if (yes_state) { | |
6470 | goto yes; | |
6471 | } else { | |
6472 | goto final_exit; | |
6473 | } | |
6474 | } | |
aa283a38 DM |
6475 | if (depth) { |
6476 | /* there's a previous state to backtrack to */ | |
40a82448 DM |
6477 | st--; |
6478 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
6479 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
6480 | st = SLAB_LAST(PL_regmatch_slab); | |
6481 | } | |
6482 | PL_regmatch_state = st; | |
40a82448 | 6483 | locinput= st->locinput; |
40a82448 | 6484 | |
5bc10b2c DM |
6485 | DEBUG_STATE_pp("pop"); |
6486 | depth--; | |
262b90c4 DM |
6487 | if (yes_state == st) |
6488 | yes_state = st->u.yes.prev_yes_state; | |
5bc10b2c | 6489 | |
24d3c4a9 DM |
6490 | state_num = st->resume_state + 1; /* failure = success + 1 */ |
6491 | goto reenter_switch; | |
95b24440 | 6492 | } |
24d3c4a9 | 6493 | result = 0; |
aa283a38 | 6494 | |
262b90c4 | 6495 | final_exit: |
bbe252da | 6496 | if (rex->intflags & PREGf_VERBARG_SEEN) { |
5d458dd8 YO |
6497 | SV *sv_err = get_sv("REGERROR", 1); |
6498 | SV *sv_mrk = get_sv("REGMARK", 1); | |
6499 | if (result) { | |
e2e6a0f1 | 6500 | sv_commit = &PL_sv_no; |
5d458dd8 YO |
6501 | if (!sv_yes_mark) |
6502 | sv_yes_mark = &PL_sv_yes; | |
6503 | } else { | |
6504 | if (!sv_commit) | |
6505 | sv_commit = &PL_sv_yes; | |
6506 | sv_yes_mark = &PL_sv_no; | |
6507 | } | |
6508 | sv_setsv(sv_err, sv_commit); | |
6509 | sv_setsv(sv_mrk, sv_yes_mark); | |
e2e6a0f1 | 6510 | } |
19b95bf0 | 6511 | |
81ed78b2 DM |
6512 | |
6513 | if (last_pushed_cv) { | |
6514 | dSP; | |
6515 | POP_MULTICALL; | |
4f8dbb2d | 6516 | PERL_UNUSED_VAR(SP); |
81ed78b2 DM |
6517 | } |
6518 | ||
2f554ef7 DM |
6519 | /* clean up; in particular, free all slabs above current one */ |
6520 | LEAVE_SCOPE(oldsave); | |
5d9a96ca | 6521 | |
730f4c74 DM |
6522 | assert(!result || locinput - PL_bostr >= 0); |
6523 | return result ? locinput - PL_bostr : -1; | |
a687059c LW |
6524 | } |
6525 | ||
6526 | /* | |
6527 | - regrepeat - repeatedly match something simple, report how many | |
d60de1d1 | 6528 | * |
e64f369d KW |
6529 | * What 'simple' means is a node which can be the operand of a quantifier like |
6530 | * '+', or {1,3} | |
6531 | * | |
d60de1d1 DM |
6532 | * startposp - pointer a pointer to the start position. This is updated |
6533 | * to point to the byte following the highest successful | |
6534 | * match. | |
6535 | * p - the regnode to be repeatedly matched against. | |
4063ade8 | 6536 | * max - maximum number of things to match. |
d60de1d1 | 6537 | * depth - (for debugging) backtracking depth. |
a687059c | 6538 | */ |
76e3520e | 6539 | STATIC I32 |
f73aaa43 | 6540 | S_regrepeat(pTHX_ const regexp *prog, char **startposp, const regnode *p, I32 max, int depth) |
a687059c | 6541 | { |
27da23d5 | 6542 | dVAR; |
4063ade8 | 6543 | char *scan; /* Pointer to current position in target string */ |
eb578fdb | 6544 | I32 c; |
4063ade8 KW |
6545 | char *loceol = PL_regeol; /* local version */ |
6546 | I32 hardcount = 0; /* How many matches so far */ | |
eb578fdb | 6547 | bool utf8_target = PL_reg_match_utf8; |
d513472c | 6548 | UV utf8_flags; |
4f55667c SP |
6549 | #ifndef DEBUGGING |
6550 | PERL_UNUSED_ARG(depth); | |
6551 | #endif | |
a0d0e21e | 6552 | |
7918f24d NC |
6553 | PERL_ARGS_ASSERT_REGREPEAT; |
6554 | ||
f73aaa43 | 6555 | scan = *startposp; |
faf11cac HS |
6556 | if (max == REG_INFTY) |
6557 | max = I32_MAX; | |
4063ade8 | 6558 | else if (! utf8_target && scan + max < loceol) |
7f596f4c | 6559 | loceol = scan + max; |
4063ade8 KW |
6560 | |
6561 | /* Here, for the case of a non-UTF-8 target we have adjusted <loceol> down | |
6562 | * to the maximum of how far we should go in it (leaving it set to the real | |
6563 | * end, if the maximum permissible would take us beyond that). This allows | |
6564 | * us to make the loop exit condition that we haven't gone past <loceol> to | |
6565 | * also mean that we haven't exceeded the max permissible count, saving a | |
6566 | * test each time through the loop. But it assumes that the OP matches a | |
6567 | * single byte, which is true for most of the OPs below when applied to a | |
6568 | * non-UTF-8 target. Those relatively few OPs that don't have this | |
6569 | * characteristic will have to compensate. | |
6570 | * | |
6571 | * There is no adjustment for UTF-8 targets, as the number of bytes per | |
6572 | * character varies. OPs will have to test both that the count is less | |
6573 | * than the max permissible (using <hardcount> to keep track), and that we | |
6574 | * are still within the bounds of the string (using <loceol>. A few OPs | |
6575 | * match a single byte no matter what the encoding. They can omit the max | |
6576 | * test if, for the UTF-8 case, they do the adjustment that was skipped | |
6577 | * above. | |
6578 | * | |
6579 | * Thus, the code above sets things up for the common case; and exceptional | |
6580 | * cases need extra work; the common case is to make sure <scan> doesn't | |
6581 | * go past <loceol>, and for UTF-8 to also use <hardcount> to make sure the | |
6582 | * count doesn't exceed the maximum permissible */ | |
6583 | ||
a0d0e21e | 6584 | switch (OP(p)) { |
22c35a8c | 6585 | case REG_ANY: |
f2ed9b32 | 6586 | if (utf8_target) { |
1aa99e6b | 6587 | while (scan < loceol && hardcount < max && *scan != '\n') { |
ffc61ed2 JH |
6588 | scan += UTF8SKIP(scan); |
6589 | hardcount++; | |
6590 | } | |
6591 | } else { | |
6592 | while (scan < loceol && *scan != '\n') | |
6593 | scan++; | |
a0ed51b3 LW |
6594 | } |
6595 | break; | |
ffc61ed2 | 6596 | case SANY: |
f2ed9b32 | 6597 | if (utf8_target) { |
a0804c9e | 6598 | while (scan < loceol && hardcount < max) { |
def8e4ea JH |
6599 | scan += UTF8SKIP(scan); |
6600 | hardcount++; | |
6601 | } | |
6602 | } | |
6603 | else | |
6604 | scan = loceol; | |
a0ed51b3 | 6605 | break; |
4063ade8 KW |
6606 | case CANY: /* Move <scan> forward <max> bytes, unless goes off end */ |
6607 | if (utf8_target && scan + max < loceol) { | |
6608 | ||
6609 | /* <loceol> hadn't been adjusted in the UTF-8 case */ | |
6610 | scan += max; | |
6611 | } | |
6612 | else { | |
6613 | scan = loceol; | |
6614 | } | |
f33976b4 | 6615 | break; |
59d32103 | 6616 | case EXACT: |
613a425d KW |
6617 | assert(STR_LEN(p) == (UTF_PATTERN) ? UTF8SKIP(STRING(p)) : 1); |
6618 | ||
59d32103 | 6619 | c = (U8)*STRING(p); |
59d32103 | 6620 | |
5e4a1da1 KW |
6621 | /* Can use a simple loop if the pattern char to match on is invariant |
6622 | * under UTF-8, or both target and pattern aren't UTF-8. Note that we | |
6623 | * can use UTF8_IS_INVARIANT() even if the pattern isn't UTF-8, as it's | |
6624 | * true iff it doesn't matter if the argument is in UTF-8 or not */ | |
6625 | if (UTF8_IS_INVARIANT(c) || (! utf8_target && ! UTF_PATTERN)) { | |
4063ade8 KW |
6626 | if (utf8_target && scan + max < loceol) { |
6627 | /* We didn't adjust <loceol> because is UTF-8, but ok to do so, | |
6628 | * since here, to match at all, 1 char == 1 byte */ | |
6629 | loceol = scan + max; | |
6630 | } | |
59d32103 KW |
6631 | while (scan < loceol && UCHARAT(scan) == c) { |
6632 | scan++; | |
6633 | } | |
6634 | } | |
b40a2c17 | 6635 | else if (UTF_PATTERN) { |
5e4a1da1 KW |
6636 | if (utf8_target) { |
6637 | STRLEN scan_char_len; | |
5e4a1da1 | 6638 | |
4063ade8 | 6639 | /* When both target and pattern are UTF-8, we have to do |
5e4a1da1 KW |
6640 | * string EQ */ |
6641 | while (hardcount < max | |
6642 | && scan + (scan_char_len = UTF8SKIP(scan)) <= loceol | |
6643 | && scan_char_len <= STR_LEN(p) | |
6644 | && memEQ(scan, STRING(p), scan_char_len)) | |
6645 | { | |
6646 | scan += scan_char_len; | |
6647 | hardcount++; | |
6648 | } | |
6649 | } | |
6650 | else if (! UTF8_IS_ABOVE_LATIN1(c)) { | |
b40a2c17 | 6651 | |
5e4a1da1 KW |
6652 | /* Target isn't utf8; convert the character in the UTF-8 |
6653 | * pattern to non-UTF8, and do a simple loop */ | |
6654 | c = TWO_BYTE_UTF8_TO_UNI(c, *(STRING(p) + 1)); | |
6655 | while (scan < loceol && UCHARAT(scan) == c) { | |
6656 | scan++; | |
6657 | } | |
6658 | } /* else pattern char is above Latin1, can't possibly match the | |
6659 | non-UTF-8 target */ | |
b40a2c17 | 6660 | } |
5e4a1da1 | 6661 | else { |
59d32103 | 6662 | |
5e4a1da1 KW |
6663 | /* Here, the string must be utf8; pattern isn't, and <c> is |
6664 | * different in utf8 than not, so can't compare them directly. | |
6665 | * Outside the loop, find the two utf8 bytes that represent c, and | |
6666 | * then look for those in sequence in the utf8 string */ | |
59d32103 KW |
6667 | U8 high = UTF8_TWO_BYTE_HI(c); |
6668 | U8 low = UTF8_TWO_BYTE_LO(c); | |
59d32103 KW |
6669 | |
6670 | while (hardcount < max | |
6671 | && scan + 1 < loceol | |
6672 | && UCHARAT(scan) == high | |
6673 | && UCHARAT(scan + 1) == low) | |
6674 | { | |
6675 | scan += 2; | |
6676 | hardcount++; | |
6677 | } | |
6678 | } | |
6679 | break; | |
5e4a1da1 | 6680 | |
2f7f8cb1 KW |
6681 | case EXACTFA: |
6682 | utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
6683 | goto do_exactf; | |
6684 | ||
d4e0b827 KW |
6685 | case EXACTFL: |
6686 | PL_reg_flags |= RF_tainted; | |
17580e7a KW |
6687 | utf8_flags = FOLDEQ_UTF8_LOCALE; |
6688 | goto do_exactf; | |
6689 | ||
d4e0b827 | 6690 | case EXACTF: |
62bf7766 KW |
6691 | utf8_flags = 0; |
6692 | goto do_exactf; | |
6693 | ||
3c760661 | 6694 | case EXACTFU_SS: |
fab2782b | 6695 | case EXACTFU_TRICKYFOLD: |
9a5a5549 | 6696 | case EXACTFU: |
05f861a2 | 6697 | utf8_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0; |
59d32103 | 6698 | |
613a425d KW |
6699 | do_exactf: { |
6700 | int c1, c2; | |
6701 | U8 c1_utf8[UTF8_MAXBYTES+1], c2_utf8[UTF8_MAXBYTES+1]; | |
d4e0b827 | 6702 | |
613a425d KW |
6703 | assert(STR_LEN(p) == (UTF_PATTERN) ? UTF8SKIP(STRING(p)) : 1); |
6704 | ||
6705 | if (S_setup_EXACTISH_ST_c1_c2(aTHX_ p, &c1, c1_utf8, &c2, c2_utf8)) { | |
6706 | if (c1 == CHRTEST_VOID) { | |
49b95fad | 6707 | /* Use full Unicode fold matching */ |
4063ade8 | 6708 | char *tmpeol = PL_regeol; |
49b95fad KW |
6709 | STRLEN pat_len = (UTF_PATTERN) ? UTF8SKIP(STRING(p)) : 1; |
6710 | while (hardcount < max | |
6711 | && foldEQ_utf8_flags(scan, &tmpeol, 0, utf8_target, | |
6712 | STRING(p), NULL, pat_len, | |
6713 | cBOOL(UTF_PATTERN), utf8_flags)) | |
6714 | { | |
6715 | scan = tmpeol; | |
4063ade8 | 6716 | tmpeol = PL_regeol; |
49b95fad KW |
6717 | hardcount++; |
6718 | } | |
613a425d KW |
6719 | } |
6720 | else if (utf8_target) { | |
6721 | if (c1 == c2) { | |
4063ade8 KW |
6722 | while (scan < loceol |
6723 | && hardcount < max | |
613a425d KW |
6724 | && memEQ(scan, c1_utf8, UTF8SKIP(scan))) |
6725 | { | |
6726 | scan += UTF8SKIP(scan); | |
6727 | hardcount++; | |
6728 | } | |
6729 | } | |
6730 | else { | |
4063ade8 KW |
6731 | while (scan < loceol |
6732 | && hardcount < max | |
613a425d KW |
6733 | && (memEQ(scan, c1_utf8, UTF8SKIP(scan)) |
6734 | || memEQ(scan, c2_utf8, UTF8SKIP(scan)))) | |
6735 | { | |
6736 | scan += UTF8SKIP(scan); | |
6737 | hardcount++; | |
6738 | } | |
6739 | } | |
6740 | } | |
6741 | else if (c1 == c2) { | |
6742 | while (scan < loceol && UCHARAT(scan) == c1) { | |
6743 | scan++; | |
6744 | } | |
6745 | } | |
6746 | else { | |
6747 | while (scan < loceol && | |
6748 | (UCHARAT(scan) == c1 || UCHARAT(scan) == c2)) | |
6749 | { | |
6750 | scan++; | |
6751 | } | |
6752 | } | |
634c83a2 | 6753 | } |
bbce6d69 | 6754 | break; |
613a425d | 6755 | } |
a0d0e21e | 6756 | case ANYOF: |
e0193e47 | 6757 | if (utf8_target) { |
4e8910e0 | 6758 | STRLEN inclasslen; |
4e8910e0 KW |
6759 | inclasslen = loceol - scan; |
6760 | while (hardcount < max | |
6761 | && ((inclasslen = loceol - scan) > 0) | |
6762 | && reginclass(prog, p, (U8*)scan, &inclasslen, utf8_target)) | |
6763 | { | |
6764 | scan += inclasslen; | |
ffc61ed2 JH |
6765 | hardcount++; |
6766 | } | |
6767 | } else { | |
32fc9b6a | 6768 | while (scan < loceol && REGINCLASS(prog, p, (U8*)scan)) |
ffc61ed2 JH |
6769 | scan++; |
6770 | } | |
a0d0e21e | 6771 | break; |
980866de | 6772 | case ALNUMU: |
f2ed9b32 | 6773 | if (utf8_target) { |
980866de | 6774 | utf8_wordchar: |
1a4fad37 | 6775 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 6776 | while (hardcount < max && scan < loceol && |
a12cf05f KW |
6777 | swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
6778 | { | |
ffc61ed2 JH |
6779 | scan += UTF8SKIP(scan); |
6780 | hardcount++; | |
6781 | } | |
980866de | 6782 | } else { |
a12cf05f KW |
6783 | while (scan < loceol && isWORDCHAR_L1((U8) *scan)) { |
6784 | scan++; | |
6785 | } | |
980866de KW |
6786 | } |
6787 | break; | |
6788 | case ALNUM: | |
6789 | if (utf8_target) | |
6790 | goto utf8_wordchar; | |
6791 | while (scan < loceol && isALNUM((U8) *scan)) { | |
6792 | scan++; | |
a0ed51b3 LW |
6793 | } |
6794 | break; | |
cfaf538b | 6795 | case ALNUMA: |
4063ade8 KW |
6796 | if (utf8_target && scan + max < loceol) { |
6797 | ||
6798 | /* We didn't adjust <loceol> because is UTF-8, but ok to do so, | |
6799 | * since here, to match, 1 char == 1 byte */ | |
6800 | loceol = scan + max; | |
6801 | } | |
cfaf538b KW |
6802 | while (scan < loceol && isWORDCHAR_A((U8) *scan)) { |
6803 | scan++; | |
6804 | } | |
6805 | break; | |
bbce6d69 | 6806 | case ALNUML: |
3280af22 | 6807 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6808 | if (utf8_target) { |
1aa99e6b IH |
6809 | while (hardcount < max && scan < loceol && |
6810 | isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
6811 | scan += UTF8SKIP(scan); |
6812 | hardcount++; | |
6813 | } | |
6814 | } else { | |
6815 | while (scan < loceol && isALNUM_LC(*scan)) | |
6816 | scan++; | |
a0ed51b3 LW |
6817 | } |
6818 | break; | |
980866de | 6819 | case NALNUMU: |
f2ed9b32 | 6820 | if (utf8_target) { |
980866de KW |
6821 | |
6822 | utf8_Nwordchar: | |
6823 | ||
1a4fad37 | 6824 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 6825 | while (hardcount < max && scan < loceol && |
980866de | 6826 | ! swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
a12cf05f | 6827 | { |
ffc61ed2 JH |
6828 | scan += UTF8SKIP(scan); |
6829 | hardcount++; | |
6830 | } | |
980866de | 6831 | } else { |
a12cf05f KW |
6832 | while (scan < loceol && ! isWORDCHAR_L1((U8) *scan)) { |
6833 | scan++; | |
6834 | } | |
980866de KW |
6835 | } |
6836 | break; | |
6837 | case NALNUM: | |
6838 | if (utf8_target) | |
6839 | goto utf8_Nwordchar; | |
6840 | while (scan < loceol && ! isALNUM((U8) *scan)) { | |
6841 | scan++; | |
a0ed51b3 LW |
6842 | } |
6843 | break; | |
0658cdde KW |
6844 | |
6845 | case POSIXA: | |
4063ade8 KW |
6846 | if (utf8_target && scan + max < loceol) { |
6847 | ||
6848 | /* We didn't adjust <loceol> because is UTF-8, but ok to do so, | |
6849 | * since here, to match, 1 char == 1 byte */ | |
6850 | loceol = scan + max; | |
6851 | } | |
6852 | while (scan < loceol && _generic_isCC_A((U8) *scan, FLAGS(p))) { | |
0658cdde KW |
6853 | scan++; |
6854 | } | |
6855 | break; | |
6856 | case NPOSIXA: | |
6857 | if (utf8_target) { | |
4063ade8 KW |
6858 | while (scan < loceol && hardcount < max |
6859 | && ! _generic_isCC_A((U8) *scan, FLAGS(p))) | |
6860 | { | |
0658cdde | 6861 | scan += UTF8SKIP(scan); |
4063ade8 | 6862 | hardcount++; |
0658cdde KW |
6863 | } |
6864 | } | |
6865 | else { | |
6866 | while (scan < loceol && ! _generic_isCC_A((U8) *scan, FLAGS(p))) { | |
6867 | scan++; | |
6868 | } | |
6869 | } | |
6870 | break; | |
cfaf538b KW |
6871 | case NALNUMA: |
6872 | if (utf8_target) { | |
4063ade8 KW |
6873 | while (scan < loceol && hardcount < max |
6874 | && ! isWORDCHAR_A((U8) *scan)) | |
6875 | { | |
cfaf538b | 6876 | scan += UTF8SKIP(scan); |
4063ade8 | 6877 | hardcount++; |
cfaf538b KW |
6878 | } |
6879 | } | |
6880 | else { | |
6881 | while (scan < loceol && ! isWORDCHAR_A((U8) *scan)) { | |
6882 | scan++; | |
6883 | } | |
6884 | } | |
6885 | break; | |
bbce6d69 | 6886 | case NALNUML: |
3280af22 | 6887 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6888 | if (utf8_target) { |
1aa99e6b IH |
6889 | while (hardcount < max && scan < loceol && |
6890 | !isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
6891 | scan += UTF8SKIP(scan); |
6892 | hardcount++; | |
6893 | } | |
6894 | } else { | |
6895 | while (scan < loceol && !isALNUM_LC(*scan)) | |
6896 | scan++; | |
a0ed51b3 LW |
6897 | } |
6898 | break; | |
980866de | 6899 | case SPACEU: |
f2ed9b32 | 6900 | if (utf8_target) { |
980866de KW |
6901 | |
6902 | utf8_space: | |
6903 | ||
1a4fad37 | 6904 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6905 | while (hardcount < max && scan < loceol && |
3568d838 | 6906 | (*scan == ' ' || |
a12cf05f KW |
6907 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) |
6908 | { | |
ffc61ed2 JH |
6909 | scan += UTF8SKIP(scan); |
6910 | hardcount++; | |
6911 | } | |
980866de KW |
6912 | break; |
6913 | } | |
6914 | else { | |
a12cf05f KW |
6915 | while (scan < loceol && isSPACE_L1((U8) *scan)) { |
6916 | scan++; | |
6917 | } | |
980866de KW |
6918 | break; |
6919 | } | |
6920 | case SPACE: | |
6921 | if (utf8_target) | |
6922 | goto utf8_space; | |
6923 | ||
6924 | while (scan < loceol && isSPACE((U8) *scan)) { | |
6925 | scan++; | |
a0ed51b3 LW |
6926 | } |
6927 | break; | |
cfaf538b | 6928 | case SPACEA: |
4063ade8 KW |
6929 | if (utf8_target && scan + max < loceol) { |
6930 | ||
6931 | /* We didn't adjust <loceol> because is UTF-8, but ok to do so, | |
6932 | * since here, to match, 1 char == 1 byte */ | |
6933 | loceol = scan + max; | |
6934 | } | |
cfaf538b KW |
6935 | while (scan < loceol && isSPACE_A((U8) *scan)) { |
6936 | scan++; | |
6937 | } | |
6938 | break; | |
bbce6d69 | 6939 | case SPACEL: |
3280af22 | 6940 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6941 | if (utf8_target) { |
1aa99e6b | 6942 | while (hardcount < max && scan < loceol && |
6bbba904 | 6943 | isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
6944 | scan += UTF8SKIP(scan); |
6945 | hardcount++; | |
6946 | } | |
6947 | } else { | |
6948 | while (scan < loceol && isSPACE_LC(*scan)) | |
6949 | scan++; | |
a0ed51b3 LW |
6950 | } |
6951 | break; | |
980866de | 6952 | case NSPACEU: |
f2ed9b32 | 6953 | if (utf8_target) { |
980866de KW |
6954 | |
6955 | utf8_Nspace: | |
6956 | ||
1a4fad37 | 6957 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6958 | while (hardcount < max && scan < loceol && |
980866de KW |
6959 | ! (*scan == ' ' || |
6960 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) | |
a12cf05f | 6961 | { |
ffc61ed2 JH |
6962 | scan += UTF8SKIP(scan); |
6963 | hardcount++; | |
6964 | } | |
980866de KW |
6965 | break; |
6966 | } | |
6967 | else { | |
a12cf05f KW |
6968 | while (scan < loceol && ! isSPACE_L1((U8) *scan)) { |
6969 | scan++; | |
6970 | } | |
980866de KW |
6971 | } |
6972 | break; | |
6973 | case NSPACE: | |
6974 | if (utf8_target) | |
6975 | goto utf8_Nspace; | |
6976 | ||
6977 | while (scan < loceol && ! isSPACE((U8) *scan)) { | |
6978 | scan++; | |
a0ed51b3 | 6979 | } |
0008a298 | 6980 | break; |
cfaf538b KW |
6981 | case NSPACEA: |
6982 | if (utf8_target) { | |
4063ade8 KW |
6983 | while (hardcount < max && scan < loceol |
6984 | && ! isSPACE_A((U8) *scan)) | |
6985 | { | |
cfaf538b | 6986 | scan += UTF8SKIP(scan); |
4063ade8 | 6987 | hardcount++; |
cfaf538b KW |
6988 | } |
6989 | } | |
6990 | else { | |
6991 | while (scan < loceol && ! isSPACE_A((U8) *scan)) { | |
6992 | scan++; | |
6993 | } | |
6994 | } | |
6995 | break; | |
bbce6d69 | 6996 | case NSPACEL: |
3280af22 | 6997 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6998 | if (utf8_target) { |
1aa99e6b | 6999 | while (hardcount < max && scan < loceol && |
6bbba904 | 7000 | !isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
7001 | scan += UTF8SKIP(scan); |
7002 | hardcount++; | |
7003 | } | |
7004 | } else { | |
7005 | while (scan < loceol && !isSPACE_LC(*scan)) | |
7006 | scan++; | |
a0ed51b3 LW |
7007 | } |
7008 | break; | |
a0d0e21e | 7009 | case DIGIT: |
f2ed9b32 | 7010 | if (utf8_target) { |
1a4fad37 | 7011 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 7012 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 7013 | swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
7014 | scan += UTF8SKIP(scan); |
7015 | hardcount++; | |
7016 | } | |
7017 | } else { | |
7018 | while (scan < loceol && isDIGIT(*scan)) | |
7019 | scan++; | |
a0ed51b3 LW |
7020 | } |
7021 | break; | |
cfaf538b | 7022 | case DIGITA: |
4063ade8 KW |
7023 | if (utf8_target && scan + max < loceol) { |
7024 | ||
7025 | /* We didn't adjust <loceol> because is UTF-8, but ok to do so, | |
7026 | * since here, to match, 1 char == 1 byte */ | |
7027 | loceol = scan + max; | |
7028 | } | |
cfaf538b KW |
7029 | while (scan < loceol && isDIGIT_A((U8) *scan)) { |
7030 | scan++; | |
7031 | } | |
7032 | break; | |
b77393f6 KW |
7033 | case DIGITL: |
7034 | PL_reg_flags |= RF_tainted; | |
7035 | if (utf8_target) { | |
b77393f6 KW |
7036 | while (hardcount < max && scan < loceol && |
7037 | isDIGIT_LC_utf8((U8*)scan)) { | |
7038 | scan += UTF8SKIP(scan); | |
7039 | hardcount++; | |
7040 | } | |
7041 | } else { | |
7042 | while (scan < loceol && isDIGIT_LC(*scan)) | |
7043 | scan++; | |
7044 | } | |
7045 | break; | |
a0d0e21e | 7046 | case NDIGIT: |
f2ed9b32 | 7047 | if (utf8_target) { |
1a4fad37 | 7048 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 7049 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 7050 | !swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
7051 | scan += UTF8SKIP(scan); |
7052 | hardcount++; | |
7053 | } | |
7054 | } else { | |
7055 | while (scan < loceol && !isDIGIT(*scan)) | |
7056 | scan++; | |
a0ed51b3 | 7057 | } |
cfaf538b KW |
7058 | break; |
7059 | case NDIGITA: | |
7060 | if (utf8_target) { | |
4063ade8 KW |
7061 | while (hardcount < max && scan < loceol |
7062 | && ! isDIGIT_A((U8) *scan)) { | |
cfaf538b | 7063 | scan += UTF8SKIP(scan); |
4063ade8 | 7064 | hardcount++; |
cfaf538b KW |
7065 | } |
7066 | } | |
7067 | else { | |
7068 | while (scan < loceol && ! isDIGIT_A((U8) *scan)) { | |
7069 | scan++; | |
7070 | } | |
7071 | } | |
7072 | break; | |
b77393f6 KW |
7073 | case NDIGITL: |
7074 | PL_reg_flags |= RF_tainted; | |
7075 | if (utf8_target) { | |
b77393f6 KW |
7076 | while (hardcount < max && scan < loceol && |
7077 | !isDIGIT_LC_utf8((U8*)scan)) { | |
7078 | scan += UTF8SKIP(scan); | |
7079 | hardcount++; | |
7080 | } | |
7081 | } else { | |
7082 | while (scan < loceol && !isDIGIT_LC(*scan)) | |
7083 | scan++; | |
7084 | } | |
7085 | break; | |
e1d1eefb | 7086 | case LNBREAK: |
e64f369d KW |
7087 | if (utf8_target) { |
7088 | while (hardcount < max && scan < loceol && | |
7089 | (c=is_LNBREAK_utf8_safe(scan, loceol))) { | |
7090 | scan += c; | |
7091 | hardcount++; | |
7092 | } | |
7093 | } else { | |
7094 | /* LNBREAK can match one or two latin chars, which is ok, but we | |
7095 | * have to use hardcount in this situation, and throw away the | |
7096 | * adjustment to <loceol> done before the switch statement */ | |
7097 | loceol = PL_regeol; | |
7098 | while (scan < loceol && (c=is_LNBREAK_latin1_safe(scan, loceol))) { | |
7099 | scan+=c; | |
7100 | hardcount++; | |
7101 | } | |
7102 | } | |
7103 | break; | |
e1d1eefb | 7104 | case HORIZWS: |
f2ed9b32 | 7105 | if (utf8_target) { |
7016d6eb DM |
7106 | while (hardcount < max && scan < loceol && |
7107 | (c=is_HORIZWS_utf8_safe(scan, loceol))) | |
7108 | { | |
e1d1eefb YO |
7109 | scan += c; |
7110 | hardcount++; | |
7111 | } | |
7112 | } else { | |
7016d6eb | 7113 | while (scan < loceol && is_HORIZWS_latin1_safe(scan, loceol)) |
e1d1eefb YO |
7114 | scan++; |
7115 | } | |
a0ed51b3 | 7116 | break; |
e1d1eefb | 7117 | case NHORIZWS: |
f2ed9b32 | 7118 | if (utf8_target) { |
7016d6eb DM |
7119 | while (hardcount < max && scan < loceol && |
7120 | !is_HORIZWS_utf8_safe(scan, loceol)) | |
7121 | { | |
e1d1eefb YO |
7122 | scan += UTF8SKIP(scan); |
7123 | hardcount++; | |
7124 | } | |
7125 | } else { | |
7016d6eb | 7126 | while (scan < loceol && !is_HORIZWS_latin1_safe(scan, loceol)) |
e1d1eefb YO |
7127 | scan++; |
7128 | ||
7129 | } | |
7130 | break; | |
7131 | case VERTWS: | |
f2ed9b32 | 7132 | if (utf8_target) { |
7016d6eb DM |
7133 | while (hardcount < max && scan < loceol && |
7134 | (c=is_VERTWS_utf8_safe(scan, loceol))) | |
7135 | { | |
e1d1eefb YO |
7136 | scan += c; |
7137 | hardcount++; | |
7138 | } | |
7139 | } else { | |
7016d6eb | 7140 | while (scan < loceol && is_VERTWS_latin1_safe(scan, loceol)) |
e1d1eefb YO |
7141 | scan++; |
7142 | ||
7143 | } | |
7144 | break; | |
7145 | case NVERTWS: | |
f2ed9b32 | 7146 | if (utf8_target) { |
7016d6eb DM |
7147 | while (hardcount < max && scan < loceol && |
7148 | !is_VERTWS_utf8_safe(scan, loceol)) | |
7149 | { | |
e1d1eefb YO |
7150 | scan += UTF8SKIP(scan); |
7151 | hardcount++; | |
7152 | } | |
7153 | } else { | |
7016d6eb | 7154 | while (scan < loceol && !is_VERTWS_latin1_safe(scan, loceol)) |
e1d1eefb YO |
7155 | scan++; |
7156 | ||
7157 | } | |
7158 | break; | |
7159 | ||
a0d0e21e LW |
7160 | default: /* Called on something of 0 width. */ |
7161 | break; /* So match right here or not at all. */ | |
7162 | } | |
a687059c | 7163 | |
a0ed51b3 LW |
7164 | if (hardcount) |
7165 | c = hardcount; | |
7166 | else | |
f73aaa43 DM |
7167 | c = scan - *startposp; |
7168 | *startposp = scan; | |
a687059c | 7169 | |
a3621e74 | 7170 | DEBUG_r({ |
e68ec53f | 7171 | GET_RE_DEBUG_FLAGS_DECL; |
be8e71aa | 7172 | DEBUG_EXECUTE_r({ |
e68ec53f YO |
7173 | SV * const prop = sv_newmortal(); |
7174 | regprop(prog, prop, p); | |
7175 | PerlIO_printf(Perl_debug_log, | |
be8e71aa | 7176 | "%*s %s can match %"IVdf" times out of %"IVdf"...\n", |
e2e6a0f1 | 7177 | REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max); |
a3621e74 | 7178 | }); |
be8e71aa | 7179 | }); |
9041c2e3 | 7180 | |
a0d0e21e | 7181 | return(c); |
a687059c LW |
7182 | } |
7183 | ||
c277df42 | 7184 | |
be8e71aa | 7185 | #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) |
c277df42 | 7186 | /* |
6c6525b8 | 7187 | - regclass_swash - prepare the utf8 swash. Wraps the shared core version to |
e0193e47 KW |
7188 | create a copy so that changes the caller makes won't change the shared one. |
7189 | If <altsvp> is non-null, will return NULL in it, for back-compat. | |
6c6525b8 | 7190 | */ |
ffc61ed2 | 7191 | SV * |
32fc9b6a | 7192 | Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp) |
ffc61ed2 | 7193 | { |
6c6525b8 | 7194 | PERL_ARGS_ASSERT_REGCLASS_SWASH; |
e0193e47 KW |
7195 | |
7196 | if (altsvp) { | |
7197 | *altsvp = NULL; | |
7198 | } | |
7199 | ||
7200 | return newSVsv(core_regclass_swash(prog, node, doinit, listsvp)); | |
6c6525b8 KW |
7201 | } |
7202 | #endif | |
7203 | ||
7204 | STATIC SV * | |
e0193e47 | 7205 | S_core_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp) |
6c6525b8 | 7206 | { |
8c9eb58f KW |
7207 | /* Returns the swash for the input 'node' in the regex 'prog'. |
7208 | * If <doinit> is true, will attempt to create the swash if not already | |
7209 | * done. | |
7210 | * If <listsvp> is non-null, will return the swash initialization string in | |
7211 | * it. | |
8c9eb58f KW |
7212 | * Tied intimately to how regcomp.c sets up the data structure */ |
7213 | ||
97aff369 | 7214 | dVAR; |
9e55ce06 JH |
7215 | SV *sw = NULL; |
7216 | SV *si = NULL; | |
7a6c6baa KW |
7217 | SV* invlist = NULL; |
7218 | ||
f8fc2ecf YO |
7219 | RXi_GET_DECL(prog,progi); |
7220 | const struct reg_data * const data = prog ? progi->data : NULL; | |
ffc61ed2 | 7221 | |
6c6525b8 | 7222 | PERL_ARGS_ASSERT_CORE_REGCLASS_SWASH; |
7918f24d | 7223 | |
ccb2541c KW |
7224 | assert(ANYOF_NONBITMAP(node)); |
7225 | ||
4f639d21 | 7226 | if (data && data->count) { |
a3b680e6 | 7227 | const U32 n = ARG(node); |
ffc61ed2 | 7228 | |
4f639d21 | 7229 | if (data->what[n] == 's') { |
ad64d0ec NC |
7230 | SV * const rv = MUTABLE_SV(data->data[n]); |
7231 | AV * const av = MUTABLE_AV(SvRV(rv)); | |
2d03de9c | 7232 | SV **const ary = AvARRAY(av); |
87367d5f | 7233 | U8 swash_init_flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; |
9041c2e3 | 7234 | |
8c9eb58f | 7235 | si = *ary; /* ary[0] = the string to initialize the swash with */ |
b11f357e | 7236 | |
88675427 KW |
7237 | /* Elements 2 and 3 are either both present or both absent. [2] is |
7238 | * any inversion list generated at compile time; [3] indicates if | |
7a6c6baa | 7239 | * that inversion list has any user-defined properties in it. */ |
88675427 KW |
7240 | if (av_len(av) >= 2) { |
7241 | invlist = ary[2]; | |
7242 | if (SvUV(ary[3])) { | |
83199d38 KW |
7243 | swash_init_flags |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY; |
7244 | } | |
7a6c6baa KW |
7245 | } |
7246 | else { | |
7247 | invlist = NULL; | |
7a6c6baa KW |
7248 | } |
7249 | ||
8c9eb58f KW |
7250 | /* Element [1] is reserved for the set-up swash. If already there, |
7251 | * return it; if not, create it and store it there */ | |
f192cf32 KW |
7252 | if (SvROK(ary[1])) { |
7253 | sw = ary[1]; | |
7254 | } | |
ffc61ed2 | 7255 | else if (si && doinit) { |
7a6c6baa KW |
7256 | |
7257 | sw = _core_swash_init("utf8", /* the utf8 package */ | |
7258 | "", /* nameless */ | |
7259 | si, | |
7260 | 1, /* binary */ | |
7261 | 0, /* not from tr/// */ | |
7a6c6baa | 7262 | invlist, |
83199d38 | 7263 | &swash_init_flags); |
ffc61ed2 JH |
7264 | (void)av_store(av, 1, sw); |
7265 | } | |
7266 | } | |
7267 | } | |
7268 | ||
7a6c6baa KW |
7269 | if (listsvp) { |
7270 | SV* matches_string = newSVpvn("", 0); | |
7a6c6baa KW |
7271 | |
7272 | /* Use the swash, if any, which has to have incorporated into it all | |
7273 | * possibilities */ | |
872dd7e0 KW |
7274 | if ((! sw || (invlist = _get_swash_invlist(sw)) == NULL) |
7275 | && (si && si != &PL_sv_undef)) | |
7276 | { | |
7a6c6baa | 7277 | |
872dd7e0 | 7278 | /* If no swash, use the input initialization string, if available */ |
7a6c6baa KW |
7279 | sv_catsv(matches_string, si); |
7280 | } | |
7281 | ||
7282 | /* Add the inversion list to whatever we have. This may have come from | |
7283 | * the swash, or from an input parameter */ | |
7284 | if (invlist) { | |
7285 | sv_catsv(matches_string, _invlist_contents(invlist)); | |
7286 | } | |
7287 | *listsvp = matches_string; | |
7288 | } | |
7289 | ||
ffc61ed2 JH |
7290 | return sw; |
7291 | } | |
7292 | ||
7293 | /* | |
ba7b4546 | 7294 | - reginclass - determine if a character falls into a character class |
832705d4 | 7295 | |
6698fab5 KW |
7296 | n is the ANYOF regnode |
7297 | p is the target string | |
7298 | lenp is pointer to the maximum number of bytes of how far to go in p | |
7299 | (This is assumed wthout checking to always be at least the current | |
7300 | character's size) | |
7301 | utf8_target tells whether p is in UTF-8. | |
832705d4 | 7302 | |
4b3cda86 KW |
7303 | Returns true if matched; false otherwise. If lenp is not NULL, on return |
7304 | from a successful match, the value it points to will be updated to how many | |
7305 | bytes in p were matched. If there was no match, the value is undefined, | |
7306 | possibly changed from the input. | |
eba1359e | 7307 | |
d5788240 KW |
7308 | Note that this can be a synthetic start class, a combination of various |
7309 | nodes, so things you think might be mutually exclusive, such as locale, | |
7310 | aren't. It can match both locale and non-locale | |
7311 | ||
bbce6d69 | 7312 | */ |
7313 | ||
76e3520e | 7314 | STATIC bool |
f6ad78d8 | 7315 | S_reginclass(pTHX_ const regexp * const prog, register const regnode * const n, register const U8* const p, STRLEN* lenp, register const bool utf8_target) |
bbce6d69 | 7316 | { |
27da23d5 | 7317 | dVAR; |
a3b680e6 | 7318 | const char flags = ANYOF_FLAGS(n); |
bbce6d69 | 7319 | bool match = FALSE; |
cc07378b | 7320 | UV c = *p; |
f7ab54c6 | 7321 | STRLEN c_len = 0; |
6698fab5 | 7322 | STRLEN maxlen; |
1aa99e6b | 7323 | |
7918f24d NC |
7324 | PERL_ARGS_ASSERT_REGINCLASS; |
7325 | ||
4b3cda86 | 7326 | /* If c is not already the code point, get it */ |
f2ed9b32 | 7327 | if (utf8_target && !UTF8_IS_INVARIANT(c)) { |
f7ab54c6 | 7328 | c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &c_len, |
6182169b KW |
7329 | (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) |
7330 | | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY); | |
7331 | /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for | |
7332 | * UTF8_ALLOW_FFFF */ | |
f7ab54c6 | 7333 | if (c_len == (STRLEN)-1) |
e8a70c6f | 7334 | Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)"); |
19f67299 | 7335 | } |
a5a291f5 KW |
7336 | else { |
7337 | c_len = 1; | |
7338 | } | |
bbce6d69 | 7339 | |
a5a291f5 KW |
7340 | /* Use passed in max length, or one character if none passed in or less |
7341 | * than one character. And assume will match just one character. This is | |
7342 | * overwritten later if matched more. */ | |
4b3cda86 | 7343 | if (lenp) { |
a5a291f5 KW |
7344 | maxlen = (*lenp > c_len) ? *lenp : c_len; |
7345 | *lenp = c_len; | |
4b3cda86 KW |
7346 | |
7347 | } | |
7348 | else { | |
a5a291f5 | 7349 | maxlen = c_len; |
4b3cda86 KW |
7350 | } |
7351 | ||
7cdde544 KW |
7352 | /* If this character is potentially in the bitmap, check it */ |
7353 | if (c < 256) { | |
ffc61ed2 JH |
7354 | if (ANYOF_BITMAP_TEST(n, c)) |
7355 | match = TRUE; | |
11454c59 KW |
7356 | else if (flags & ANYOF_NON_UTF8_LATIN1_ALL |
7357 | && ! utf8_target | |
7358 | && ! isASCII(c)) | |
7359 | { | |
7360 | match = TRUE; | |
7361 | } | |
78969a98 KW |
7362 | else if (flags & ANYOF_LOCALE) { |
7363 | PL_reg_flags |= RF_tainted; | |
7364 | ||
538b546e | 7365 | if ((flags & ANYOF_LOC_FOLD) |
78969a98 KW |
7366 | && ANYOF_BITMAP_TEST(n, PL_fold_locale[c])) |
7367 | { | |
ffc61ed2 | 7368 | match = TRUE; |
78969a98 | 7369 | } |
040aea3a KW |
7370 | else if (ANYOF_CLASS_TEST_ANY_SET(n) && |
7371 | ((ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) || | |
7372 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) || | |
7373 | (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) || | |
7374 | (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) || | |
7375 | (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) || | |
7376 | (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) || | |
7377 | (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) || | |
7378 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) || | |
7379 | (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) || | |
7380 | (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) || | |
07315176 KW |
7381 | (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII_LC(c)) || |
7382 | (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII_LC(c)) || | |
040aea3a KW |
7383 | (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) || |
7384 | (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) || | |
7385 | (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) || | |
7386 | (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) || | |
7387 | (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) || | |
7388 | (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) || | |
7389 | (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) || | |
7390 | (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) || | |
7391 | (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) || | |
7392 | (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) || | |
7393 | (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) || | |
7394 | (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) || | |
7395 | (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) || | |
7396 | (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) || | |
7397 | (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) || | |
7398 | (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) || | |
67addccf KW |
7399 | (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK_LC(c)) || |
7400 | (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK_LC(c)) | |
040aea3a | 7401 | ) /* How's that for a conditional? */ |
78969a98 | 7402 | ) { |
ffc61ed2 JH |
7403 | match = TRUE; |
7404 | } | |
a0ed51b3 | 7405 | } |
a0ed51b3 LW |
7406 | } |
7407 | ||
7cdde544 | 7408 | /* If the bitmap didn't (or couldn't) match, and something outside the |
1f327b5e | 7409 | * bitmap could match, try that. Locale nodes specify completely the |
de87c4fe | 7410 | * behavior of code points in the bit map (otherwise, a utf8 target would |
c613755a | 7411 | * cause them to be treated as Unicode and not locale), except in |
de87c4fe | 7412 | * the very unlikely event when this node is a synthetic start class, which |
c613755a KW |
7413 | * could be a combination of locale and non-locale nodes. So allow locale |
7414 | * to match for the synthetic start class, which will give a false | |
7415 | * positive that will be resolved when the match is done again as not part | |
7416 | * of the synthetic start class */ | |
ef87b810 | 7417 | if (!match) { |
10ee90d2 KW |
7418 | if (utf8_target && (flags & ANYOF_UNICODE_ALL) && c >= 256) { |
7419 | match = TRUE; /* Everything above 255 matches */ | |
e051a21d | 7420 | } |
6f8d7d0d KW |
7421 | else if (ANYOF_NONBITMAP(n) |
7422 | && ((flags & ANYOF_NONBITMAP_NON_UTF8) | |
7423 | || (utf8_target | |
7424 | && (c >=256 | |
7425 | || (! (flags & ANYOF_LOCALE)) | |
7426 | || (flags & ANYOF_IS_SYNTHETIC))))) | |
ef87b810 | 7427 | { |
e0193e47 | 7428 | SV * const sw = core_regclass_swash(prog, n, TRUE, 0); |
7cdde544 KW |
7429 | if (sw) { |
7430 | U8 * utf8_p; | |
7431 | if (utf8_target) { | |
7432 | utf8_p = (U8 *) p; | |
e0193e47 KW |
7433 | } else { /* Convert to utf8 */ |
7434 | STRLEN len = 1; | |
7cdde544 KW |
7435 | utf8_p = bytes_to_utf8(p, &len); |
7436 | } | |
f56b6394 | 7437 | |
e0193e47 | 7438 | if (swash_fetch(sw, utf8_p, TRUE)) { |
7cdde544 | 7439 | match = TRUE; |
e0193e47 | 7440 | } |
7cdde544 KW |
7441 | |
7442 | /* If we allocated a string above, free it */ | |
7443 | if (! utf8_target) Safefree(utf8_p); | |
7444 | } | |
7445 | } | |
5073ffbd KW |
7446 | |
7447 | if (UNICODE_IS_SUPER(c) | |
7448 | && (flags & ANYOF_WARN_SUPER) | |
7449 | && ckWARN_d(WARN_NON_UNICODE)) | |
7450 | { | |
7451 | Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE), | |
7452 | "Code point 0x%04"UVXf" is not Unicode, all \\p{} matches fail; all \\P{} matches succeed", c); | |
7453 | } | |
7cdde544 KW |
7454 | } |
7455 | ||
f0fdc1c9 KW |
7456 | /* The xor complements the return if to invert: 1^1 = 0, 1^0 = 1 */ |
7457 | return cBOOL(flags & ANYOF_INVERT) ^ match; | |
a0ed51b3 | 7458 | } |
161b471a | 7459 | |
dfe13c55 | 7460 | STATIC U8 * |
0ce71af7 | 7461 | S_reghop3(U8 *s, I32 off, const U8* lim) |
9041c2e3 | 7462 | { |
6af86488 KW |
7463 | /* return the position 'off' UTF-8 characters away from 's', forward if |
7464 | * 'off' >= 0, backwards if negative. But don't go outside of position | |
7465 | * 'lim', which better be < s if off < 0 */ | |
7466 | ||
97aff369 | 7467 | dVAR; |
7918f24d NC |
7468 | |
7469 | PERL_ARGS_ASSERT_REGHOP3; | |
7470 | ||
a0ed51b3 | 7471 | if (off >= 0) { |
1aa99e6b | 7472 | while (off-- && s < lim) { |
ffc61ed2 | 7473 | /* XXX could check well-formedness here */ |
a0ed51b3 | 7474 | s += UTF8SKIP(s); |
ffc61ed2 | 7475 | } |
a0ed51b3 LW |
7476 | } |
7477 | else { | |
1de06328 YO |
7478 | while (off++ && s > lim) { |
7479 | s--; | |
7480 | if (UTF8_IS_CONTINUED(*s)) { | |
7481 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
7482 | s--; | |
a0ed51b3 | 7483 | } |
1de06328 | 7484 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
7485 | } |
7486 | } | |
7487 | return s; | |
7488 | } | |
161b471a | 7489 | |
f9f4320a YO |
7490 | #ifdef XXX_dmq |
7491 | /* there are a bunch of places where we use two reghop3's that should | |
7492 | be replaced with this routine. but since thats not done yet | |
7493 | we ifdef it out - dmq | |
7494 | */ | |
dfe13c55 | 7495 | STATIC U8 * |
1de06328 YO |
7496 | S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim) |
7497 | { | |
7498 | dVAR; | |
7918f24d NC |
7499 | |
7500 | PERL_ARGS_ASSERT_REGHOP4; | |
7501 | ||
1de06328 YO |
7502 | if (off >= 0) { |
7503 | while (off-- && s < rlim) { | |
7504 | /* XXX could check well-formedness here */ | |
7505 | s += UTF8SKIP(s); | |
7506 | } | |
7507 | } | |
7508 | else { | |
7509 | while (off++ && s > llim) { | |
7510 | s--; | |
7511 | if (UTF8_IS_CONTINUED(*s)) { | |
7512 | while (s > llim && UTF8_IS_CONTINUATION(*s)) | |
7513 | s--; | |
7514 | } | |
7515 | /* XXX could check well-formedness here */ | |
7516 | } | |
7517 | } | |
7518 | return s; | |
7519 | } | |
f9f4320a | 7520 | #endif |
1de06328 YO |
7521 | |
7522 | STATIC U8 * | |
0ce71af7 | 7523 | S_reghopmaybe3(U8* s, I32 off, const U8* lim) |
a0ed51b3 | 7524 | { |
97aff369 | 7525 | dVAR; |
7918f24d NC |
7526 | |
7527 | PERL_ARGS_ASSERT_REGHOPMAYBE3; | |
7528 | ||
a0ed51b3 | 7529 | if (off >= 0) { |
1aa99e6b | 7530 | while (off-- && s < lim) { |
ffc61ed2 | 7531 | /* XXX could check well-formedness here */ |
a0ed51b3 | 7532 | s += UTF8SKIP(s); |
ffc61ed2 | 7533 | } |
a0ed51b3 | 7534 | if (off >= 0) |
3dab1dad | 7535 | return NULL; |
a0ed51b3 LW |
7536 | } |
7537 | else { | |
1de06328 YO |
7538 | while (off++ && s > lim) { |
7539 | s--; | |
7540 | if (UTF8_IS_CONTINUED(*s)) { | |
7541 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
7542 | s--; | |
a0ed51b3 | 7543 | } |
1de06328 | 7544 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
7545 | } |
7546 | if (off <= 0) | |
3dab1dad | 7547 | return NULL; |
a0ed51b3 LW |
7548 | } |
7549 | return s; | |
7550 | } | |
51371543 | 7551 | |
51371543 | 7552 | static void |
acfe0abc | 7553 | restore_pos(pTHX_ void *arg) |
51371543 | 7554 | { |
97aff369 | 7555 | dVAR; |
097eb12c | 7556 | regexp * const rex = (regexp *)arg; |
ed301438 | 7557 | if (PL_reg_state.re_state_eval_setup_done) { |
51371543 | 7558 | if (PL_reg_oldsaved) { |
4f639d21 DM |
7559 | rex->subbeg = PL_reg_oldsaved; |
7560 | rex->sublen = PL_reg_oldsavedlen; | |
6502e081 DM |
7561 | rex->suboffset = PL_reg_oldsavedoffset; |
7562 | rex->subcoffset = PL_reg_oldsavedcoffset; | |
f8c7b90f | 7563 | #ifdef PERL_OLD_COPY_ON_WRITE |
4f639d21 | 7564 | rex->saved_copy = PL_nrs; |
ed252734 | 7565 | #endif |
07bc277f | 7566 | RXp_MATCH_COPIED_on(rex); |
51371543 GS |
7567 | } |
7568 | PL_reg_magic->mg_len = PL_reg_oldpos; | |
ed301438 | 7569 | PL_reg_state.re_state_eval_setup_done = FALSE; |
51371543 GS |
7570 | PL_curpm = PL_reg_oldcurpm; |
7571 | } | |
7572 | } | |
33b8afdf JH |
7573 | |
7574 | STATIC void | |
7575 | S_to_utf8_substr(pTHX_ register regexp *prog) | |
7576 | { | |
7e0d5ad7 KW |
7577 | /* Converts substr fields in prog from bytes to UTF-8, calling fbm_compile |
7578 | * on the converted value */ | |
7579 | ||
a1cac82e | 7580 | int i = 1; |
7918f24d NC |
7581 | |
7582 | PERL_ARGS_ASSERT_TO_UTF8_SUBSTR; | |
7583 | ||
a1cac82e NC |
7584 | do { |
7585 | if (prog->substrs->data[i].substr | |
7586 | && !prog->substrs->data[i].utf8_substr) { | |
7587 | SV* const sv = newSVsv(prog->substrs->data[i].substr); | |
7588 | prog->substrs->data[i].utf8_substr = sv; | |
7589 | sv_utf8_upgrade(sv); | |
610460f9 | 7590 | if (SvVALID(prog->substrs->data[i].substr)) { |
cffe132d | 7591 | if (SvTAIL(prog->substrs->data[i].substr)) { |
610460f9 NC |
7592 | /* Trim the trailing \n that fbm_compile added last |
7593 | time. */ | |
7594 | SvCUR_set(sv, SvCUR(sv) - 1); | |
7595 | /* Whilst this makes the SV technically "invalid" (as its | |
7596 | buffer is no longer followed by "\0") when fbm_compile() | |
7597 | adds the "\n" back, a "\0" is restored. */ | |
cffe132d NC |
7598 | fbm_compile(sv, FBMcf_TAIL); |
7599 | } else | |
7600 | fbm_compile(sv, 0); | |
610460f9 | 7601 | } |
a1cac82e NC |
7602 | if (prog->substrs->data[i].substr == prog->check_substr) |
7603 | prog->check_utf8 = sv; | |
7604 | } | |
7605 | } while (i--); | |
33b8afdf JH |
7606 | } |
7607 | ||
7e0d5ad7 | 7608 | STATIC bool |
33b8afdf JH |
7609 | S_to_byte_substr(pTHX_ register regexp *prog) |
7610 | { | |
7e0d5ad7 KW |
7611 | /* Converts substr fields in prog from UTF-8 to bytes, calling fbm_compile |
7612 | * on the converted value; returns FALSE if can't be converted. */ | |
7613 | ||
97aff369 | 7614 | dVAR; |
a1cac82e | 7615 | int i = 1; |
7918f24d NC |
7616 | |
7617 | PERL_ARGS_ASSERT_TO_BYTE_SUBSTR; | |
7618 | ||
a1cac82e NC |
7619 | do { |
7620 | if (prog->substrs->data[i].utf8_substr | |
7621 | && !prog->substrs->data[i].substr) { | |
7622 | SV* sv = newSVsv(prog->substrs->data[i].utf8_substr); | |
7e0d5ad7 KW |
7623 | if (! sv_utf8_downgrade(sv, TRUE)) { |
7624 | return FALSE; | |
7625 | } | |
5400f398 KW |
7626 | if (SvVALID(prog->substrs->data[i].utf8_substr)) { |
7627 | if (SvTAIL(prog->substrs->data[i].utf8_substr)) { | |
7628 | /* Trim the trailing \n that fbm_compile added last | |
7629 | time. */ | |
7630 | SvCUR_set(sv, SvCUR(sv) - 1); | |
7631 | fbm_compile(sv, FBMcf_TAIL); | |
7632 | } else | |
7633 | fbm_compile(sv, 0); | |
7634 | } | |
a1cac82e NC |
7635 | prog->substrs->data[i].substr = sv; |
7636 | if (prog->substrs->data[i].utf8_substr == prog->check_utf8) | |
7637 | prog->check_substr = sv; | |
33b8afdf | 7638 | } |
a1cac82e | 7639 | } while (i--); |
7e0d5ad7 KW |
7640 | |
7641 | return TRUE; | |
33b8afdf | 7642 | } |
66610fdd | 7643 | |
a0e786e5 KW |
7644 | /* These constants are for finding GCB=LV and GCB=LVT. These are for the |
7645 | * pre-composed Hangul syllables, which are all in a contiguous block and | |
7646 | * arranged there in such a way so as to facilitate alorithmic determination of | |
7647 | * their characteristics. As such, they don't need a swash, but can be | |
7648 | * determined by simple arithmetic. Almost all are GCB=LVT, but every 28th one | |
7649 | * is a GCB=LV */ | |
7650 | #define SBASE 0xAC00 /* Start of block */ | |
7651 | #define SCount 11172 /* Length of block */ | |
7652 | #define TCount 28 | |
7653 | ||
7654 | #if 0 /* This routine is not currently used */ | |
7655 | PERL_STATIC_INLINE bool | |
7656 | S_is_utf8_X_LV(pTHX_ const U8 *p) | |
7657 | { | |
7658 | /* Unlike most other similarly named routines here, this does not create a | |
7659 | * swash, so swash_fetch() cannot be used on PL_utf8_X_LV. */ | |
7660 | ||
7661 | dVAR; | |
7662 | ||
7663 | UV cp = valid_utf8_to_uvchr(p, NULL); | |
7664 | ||
7665 | PERL_ARGS_ASSERT_IS_UTF8_X_LV; | |
7666 | ||
7667 | /* The earliest Unicode releases did not have these precomposed Hangul | |
7668 | * syllables. Set to point to undef in that case, so will return false on | |
7669 | * every call */ | |
7670 | if (! PL_utf8_X_LV) { /* Set up if this is the first time called */ | |
7671 | PL_utf8_X_LV = swash_init("utf8", "_X_GCB_LV", &PL_sv_undef, 1, 0); | |
7672 | if (_invlist_len(_get_swash_invlist(PL_utf8_X_LV)) == 0) { | |
7673 | SvREFCNT_dec(PL_utf8_X_LV); | |
7674 | PL_utf8_X_LV = &PL_sv_undef; | |
7675 | } | |
7676 | } | |
7677 | ||
7678 | return (PL_utf8_X_LV != &PL_sv_undef | |
7679 | && cp >= SBASE && cp < SBASE + SCount | |
7680 | && (cp - SBASE) % TCount == 0); /* Only every TCount one is LV */ | |
7681 | } | |
7682 | #endif | |
7683 | ||
7684 | PERL_STATIC_INLINE bool | |
7685 | S_is_utf8_X_LVT(pTHX_ const U8 *p) | |
7686 | { | |
7687 | /* Unlike most other similarly named routines here, this does not create a | |
7688 | * swash, so swash_fetch() cannot be used on PL_utf8_X_LVT. */ | |
7689 | ||
7690 | dVAR; | |
7691 | ||
7692 | UV cp = valid_utf8_to_uvchr(p, NULL); | |
7693 | ||
7694 | PERL_ARGS_ASSERT_IS_UTF8_X_LVT; | |
7695 | ||
7696 | /* The earliest Unicode releases did not have these precomposed Hangul | |
7697 | * syllables. Set to point to undef in that case, so will return false on | |
7698 | * every call */ | |
7699 | if (! PL_utf8_X_LVT) { /* Set up if this is the first time called */ | |
7700 | PL_utf8_X_LVT = swash_init("utf8", "_X_GCB_LVT", &PL_sv_undef, 1, 0); | |
7701 | if (_invlist_len(_get_swash_invlist(PL_utf8_X_LVT)) == 0) { | |
7702 | SvREFCNT_dec(PL_utf8_X_LVT); | |
7703 | PL_utf8_X_LVT = &PL_sv_undef; | |
7704 | } | |
7705 | } | |
7706 | ||
7707 | return (PL_utf8_X_LVT != &PL_sv_undef | |
7708 | && cp >= SBASE && cp < SBASE + SCount | |
7709 | && (cp - SBASE) % TCount != 0); /* All but every TCount one is LV */ | |
7710 | } | |
7711 | ||
66610fdd RGS |
7712 | /* |
7713 | * Local variables: | |
7714 | * c-indentation-style: bsd | |
7715 | * c-basic-offset: 4 | |
14d04a33 | 7716 | * indent-tabs-mode: nil |
66610fdd RGS |
7717 | * End: |
7718 | * | |
14d04a33 | 7719 | * ex: set ts=8 sts=4 sw=4 et: |
37442d52 | 7720 | */ |